Defining Grapheme-Color Synesthesia
RESEARCH MASTER’S PSYCHOLOGY THESIS
Graduate School of Psychology
Information
Name : Sam H. Beekhuizen
Student ID number : 6061052
Specialization : Brain & Cognition Research center : University of Amsterdam
Supervisor : Romke Rouw
Second assessor : Steven Scholte
Date : April 2016
CONTENTS
Abstract………... 3 Introduction……… 4 Methods……….. 8 Participants……….. 8 Procedure……….8 Materials……….10 Data Analysis………. 13 Results………15 Confirmatory analyses……….15 Stroop effect……… 15 Correlation………15 ‘Projectors’ vs ‘Associators’………...17 Exploratory analyses……….. 18 Discussion………. 22 Conclusion……… 22 Limitations……… 23 Future directions………... 25 References………. 27Abstract
This study compares different defining measures of the neurobiological condition known as grapheme-color synesthesia (GCS). We aim to see how these measures correlate with each other; and to gain insight into the way this type of synesthesia can be diagnosed. The goal is to generate a more coherent and realistic view of what it means to be a GC synesthete by looking at the relationships among these defining criteria. 33 grapheme-color synesthetes participated in the current study. The defining dimensions under investigation are (1) automaticity (Stroop effect), (2) consistency (Color Picker Test), (3) intensity of the synesthetic experience, and (4) the spatial location of the experienced synesthetic color (being a ‘projector’ or ‘associator’). We found a synesthetic Stroop effect for all our participants. Our defining criteria did not correlate with each other positively, but rather showed a mixed pattern of relationships. Lastly, not all variables were more pronounced for ‘projectors’ than ‘associators’. Additional patterns in the data were
investigated exploratory. The results shed light on how these dimensions of synesthesia might be related to each other.
Keywords: synesthesia, grapheme-color synesthesia, Stroop effect, ‘projector’ vs
INTRODUCTION
Research into synesthesia is a relatively young field. It was first investigated by Georg Tobias Ludwig Sachs in 1812 (see Jewanski, Day, & Ward, 2009; cited in Simner, 2012) and since then a large number of different forms of this condition have been characterized (Simner, 2012). What all forms of synesthesia have in common is that the sensory experience in one sense (the inducing experience) involuntarily leads to a perceptual experience in another sense (the concurrent experience). However, the concurrent doesn’t always have to be an experience in another sense. It can also be in the same sensory modality or of a cognitive nature. Historically synesthesia has been characterized in a number of different ways. This, in combination with the fact that synesthesia is a general term for a large number of perceptual conditions, makes it very difficult to find a definition for such a broad and ill-understood phenomenon, as pointed out by Simner (2012).
A review of the literature shows that a definition of synesthesia usually entails at least one of the following defining criteria (See Colizoli, Murre, & Rouw, 2014 and Simner, 2012 for an overview). Firstly, the consistency of the coupling between the inducing sensory stimulus and the following perceptual concurrent (Eagleman, 2007). The more consistent the binding of these experiences is, the stronger someone’s synesthesia is considered to be. Secondly, the automaticity with which the concurrent is evoked (Van der Veen, 2014). The more automatic and fast the synesthetic experience follows a ‘normal’ sensory experience the stronger the synesthesia is considered to be. Thirdly, the intensity of the synesthetic experience. This is a very subjective measure and is taken to indicate to what extent the synesthesia entails real-world properties and mimics the directness and intensity of ‘true’ sensory experiences. Fourthly, the spatial location of the synesthetic experience: is it experienced as just an association in the mind’s eye or an actual perceptual experience coming from the outside world i.e. spatially extended (Cytowic, 2002). The latter are named ‘projector’ synesthetes (Dixon et al., 2004; Prize & Mentzoni, 2008), whereas the former are named ‘associator’ synesthetes (Rouw et al., 2013; Colizoli et al., 2014).
In recent years synesthesia research has received a new surge of interest from the scientific community (Simner, 2012). However, different studies investigating synesthesia have used the previously mentioned and still other defining criteria interchangeably. It is thus difficult to compare results across studies and be confident that their results are pertaining to the same types of synesthesia. Generalizability is therefore a big problem. Usually the consistency criterion is used as a first benchmark to ‘diagnose’ someone as a synesthete. But, as Simner (2012) points out, this leads to a self-reinforcing bias: ‘synesthetes’ are selected based on a consistency criterion, and then said to be ‘synesthetes’ because they are consistent in their experiences. Thus, there should be a more coherent way of determining when someone is a synesthete and when someone isn’t. But, more importantly, do these varied defining measures of synesthesia agree with each other or do they seem to measure different dimensions of the same phenomenon?
In the current study the focus will be on grapheme-color synesthesia (GCS) since this is the form most often studied and most easily assessed through research. The prevalence of grapheme-color synesthesia is about one percent in the general population (Colizoli et al., 2014). People with grapheme-color synesthesia experience colors in response to specific letters and numbers. For example, one might experience the letter A as being red and the letter B as being blue (either on paper (‘projector’) or in the mind’s eye (‘associator’)). It is still unclear if these associations are caused by the actual shape of the graphemes or by a more semantic-like
association. Either way, people with this type of synesthesia are readily available and it is relatively easy to operationalize ways of investigating this form.
The goal of this study is to investigate how the different ways to determine GCS relate to each other. This will be done by measuring consistency (along with a measure of accuracy and visual imagery), automaticity, spatial location and subjective experience of GCS within
synesthetes. These measures will then be correlated with each other. Secondly we will look at the difference between ‘projectors’ and ‘associators’ assuming this distinction represent a meaningful difference between the two groups of synesthetes. The main question therefore is: How do
results of different defining measures of synesthesia correlate with each other? Do they co-vary in the expected directions or not? And is it true that ‘projectors’ score higher on these measures than ‘associators’?
Based on what is known and assumed about synesthesia, this leads to the following hypotheses: First, that there is a Stroop effect (automaticity) to be found with synesthetes (Van der Veen et al., 2014)(especially for those meeting the consistency criterion measure of < 1 (Eagleman Synesthesia Battery, Eagleman et al., 2007)) (H1). For this purpose a synesthetic Stroop task was used (Colizoli et al., 2014; Mattingley et al., 2001; Rouw et al., 2013; Veen et al., 2014). In a normal Stroop task color words are presented to participants in specific colors (Stroop, 1935). These colors can be either congruent or incongruent with the spelled color word. For example, RED would be an incongruent trial whereas GREEN would be a congruent trial. Participants have to respond as fast and accurately as possible to the typeface color and not to the word. Incongruent trials usually show significantly higher reaction times than congruent trials. The same principle can be used to determine synesthesia, by using graphemes presented in specific colors, either congruent or incongruent with the participant’s synesthetic colors (Rouw et al., 2013). This synesthetic Stroop effect relates to the automaticity of synesthesia; the fact that it can’t be voluntarily inhibited or ignored (Mattingley et al., 2001). It is important to note that the Stroop effect is not consistently found throughout the literature (MacLeod, 1991).
The synesthetic Stroop effect has been previously established for GCS by multiple studies (Rouw et al., 2013; Colizoli et al., 2014; Veen et al., 2014; Lupiáñez & Callejas, 2006). This is taken as evidence for the existence of this phenomenon. Another way to determine GCS has been to focus on consistency of the synesthetic experience by using ‘The Synesthesia Battery’ (Eagleman et al., 2007; Carmichael et al., 2015). Since these measures are both used to define GCS we expect them to be positively related to each other. It follows intuitively that the intensity of the synesthetic experience and the resemblance to a ‘real’ perception are positively related to higher consistency and bigger Stroop interference. Our second hypothesis is therefore that all
four measurements taken will correlate positively with each other (H2). Thus higher consistency scores relate to bigger Stroop effects, a more intense subjective synesthetic experience and a higher ‘projector/associator’ score.
Lastly, we expect that the hypothesized effects/measurements will be more pronounced in ‘projectors’ compared to ‘associators’ (H3). For instance, Dixon et al. (2004) found that ‘projectors’ showed a significantly larger synesthetic Stroop effect than ‘associators’. Ward, Salih, & Sagiv (2007) also found that ‘projectors’ are faster at naming their synesthetic colors and experience them more intensely. Finally, there will be room for exploratory investigation into other possible covariates, such as the demographic information of the synesthetes (age, sex, number of forms of synesthesia, medical history).
To conclude, the field of synesthesia seems somewhat cluttered and should be cleaned up. As Simner (2012) said: ‘inherent in any broad sustained interest is the importance of
establishing a clear definition of the focus of study, although remarkably, the literature contains a number of conflicting assumptions about the very definition of synesthesia.’ This study is thus a project that will focus on maintenance with the following main question: how do the different defining measures of GCS synesthesia relate to each other and is there a pattern to be found? The interpretation if the results match our expectations is that synesthesia seems to be a coherent and directional condition, where our measurements are positively correlated, meaning that the more intense the experience, the higher the consistency, the greater the interference effect of the Stroop task, and the greater the chance someone is a ‘projector’. This would mean that studies focusing on different aspects of synesthesia could actually be focusing on the same underlying dimension (GCS). This would mean earlier research is actually comparable.
METHODS
Participants
33 GC synesthetes (5 male, 28 female, 20 – 61 years of age (mean age: 32), 29 Dutch, 1 German, 1 French, 1 Brazilian, 1 Spanish) participated in the current study. This sample size was determined by the availability of appropriate synesthetic subjects. For two participants the consistency score was missing, and thus they were excluded pairwise from analyses involving this variable. Synesthetic participants were recruited through an available online database (Eagleman Test Battery, synesthete.org) or had previously participated in synesthesia experiments in our lab. Participants with abnormal (color) vision were excluded from this study. Scores on trials
exceeding ± 3 SD of the mean on the dependent variable (Stroop task) were also excluded from the analysis. Informed consent was acquired prior to conducting the research. Participants were rewarded with 12.50 euro for their participation.
Procedure
To obtain our data on GCS four different tests were administered to our participants: two measures were objective assessments of synesthesia: the Stroop task and the Eagleman Synesthesia Battery (Eagleman et al., 2007). The other measures were subjective in their
assessment, relying on self-report from the synesthetes: the intensity questionnaire (INT, Rouw, unpublished) and the P/A questionnaire (Rouw & Scholte, 2007). These four respectively measured the automaticity, consistency (plus accuracy and visual imagery ability), intensity and spatial location of the synesthetic percepts (see below under materials).
GC synesthetes were recruited through a pre-existing database. Furthermore, the Eagleman Synesthesia Battery is an online test that is easily administered to a large sample of people. After sending the link for this test to many people, the consistency scores on this test were used to further locate potential synesthetes (since the scores on the other tests will be correlated with the consistency scores, the self-reinforcing bias of focusing on consistency alone
(mentioned above) does not apply here). The procedures thus consisted of one session online (the Eagleman Synesthesia Battery) and one session in person (the Stroop task and the two questionnaires, completed in the lab at the University of Amsterdam). The Eagleman test battery (synesthete.org) consists of a collection of tests and measures that collect data on different kinds of synesthesia. This site functions as a database for synesthesia research and is used for many different research projects. For our purposes we only used the results from the Grapheme Color Picker Test (a test that links colors to specific graphemes and creates a consistency score) and, where available, the results of the Speedy Congruency Test (accuracy) and the Vividness of Visual Imagery (VVIQ-2 score). Depending on how many different kinds of synesthesia someone has, this online could take as long as 30 min. to two hours.
After finishing the online Eagleman test participants were invited to the University of Amsterdam to take the other tests. They were first explained the procedure and order of testing, after which they signed an informed consent form, stating that they could quit whenever they wanted and being assured absolute anonymity. They then sat with the researcher discussing which four letters/numbers and corresponding colors to use for the synesthetic Stroop task (always selecting the letters/numbers which had the most pronounced synesthetic character to that specific synesthete). They then moved on to a practice task where they learned which response buttons corresponded to which colors. This consisted of 96 trials in which the
participants were presented with colored squares with one of the four chosen colors and had to press one of four labeled keys on a USB keyboard as fast and accurately as possible. They received feedback on their performance so that they could familiarize themselves with the task.
Afterwards the Stroop task was completed (15 min.) in which the colors were presented as letters instead of squares. It consisted of 432 trials (with 30 practice trials with feedback beforehand) divided into 6 blocks of 72 trials with short breaks to regain concentration in between. Half of all trials were congruent with the synesthete’s synesthetic experience, the other half incongruent. Participants were instructed to sit upright and respond to the ‘objective’ color
in which the letters were presented as fast and accurately as possible. First a black grapheme was presented for 100 ms after which it took on its color. Participants then had 3 sec. to react before the next trial began. The Stroop primes were programmed to be 4 at 4 cm, against a white background (color: 255,255,255). The letters used as stimuli were in arial fond (colors based on subject specific synesthesia), in fond size of 110. The participants sat opposite to the computer at a distance of about 75 centimeters.
Following the Stroop task were the two questionnaires (P/A and INT, 15 min.). After completing the Stroop task and the two questionnaires the participants were thanked for their contribution and were paid 12.50 euros. They were finally explained the purpose behind the study and were told that they will receive the final paper/results on request. The testing session held at the University of Amsterdam took roughly one hour.
Materials
The following tests were used in our battery: first, a test of consistency, the Color-Picker-Test (CPT), with the possible addition of the Speeded Congruency Color-Picker-Test (SCT) and a test of Vividness of Visual Imagery (VVIQ-2), all three components of the Eagleman Synesthesia Battery, which is accessible via the internet (www.synesthete.org; Eagleman et al., 2007)). These tests respectively measure the consistency in pairing graphemes with certain colors (in RGB values), the accuracy of their responses with mean RT’s and how vividly people can visualize objects and scenes (mental imagery). During the CPT the participants had to connect their synesthetic color experiences to specific graphemes over the course of 108 trials from which consistency was then automatically calculated. Carmichael et al. (2015) tested and confirmed that ‘The Synesthesia Battery’ is a valid test to determine grapheme-color synesthesia. An important note to make here is that consistency is higher when the score is lower, so it is actually reverse coded. This is important for the interpretation of the correlations (see below). Figure 1 shows how the online consistency data are represented after participants finished the test:
Figure 1. A random participant’s result on the CPT.
Second, we used a test of automaticity (the Stroop test (adapted to synesthesia)). Since the synesthetic sensations supplement, but do not replace, the usual modality-specific perceptions (Simner, 2012) the application of a Stroop task seemed appropriate. We expected to find a Stroop effect for our synesthetes, as was found in earlier studies (Lupiáñez & Callejas, 2006). We used an online color picker tool where participants could select colors and their corresponding RGB values to adjust the Stroop task to each of our individual synesthetes’ color letter pairings (www.virtuosoft.eu/code/jquery-colorpickersliders). As mentioned above, before starting the Stroop task participants completed a practice task to connect the response buttons with the right colors. This practice task was repeated until the number of errors was below threshold (<10 mistakes). To make sure our Stroop task would work we ran a successful pilot test on a synesthetic subject.
Third, we used an Intensity questionnaire (INT, Rouw, unpublished). This is a two-part questionnaire with the first part pertaining to how much the synesthetic perception resembles a ‘real’ perception (i.e. something objectively present in the outside world) and the second part measuring how ‘sharp’, ‘clear’ and ‘powerful’ the synesthetic percept is in comparison to a ‘real’ percept (HSK). The ‘reality’ questionnaire consisted of 8 items (α = .84), with five-point Likert scale answer possibilities ranging from 1 ‘strongly disagree’ to 5 ‘strongly agree’. An example question is: ‘sometimes I confuse the synesthetic color experience with seeing a ‘real’ color (for example: it seems that the letter I am looking at really has the ‘synesthetic’ color for a moment).’ The ‘Sharpness, Clarity, powerful’ questionnaire (HSK) consisted of only three items (α = .74), contrasting a synesthetic percept with a normal one. All three questions were phrased as follows: ‘what is more clear, a synesthetic color or a real color?’ with five-point Likert scale answer possibilities ranging from -2 ‘the real color is definitely more clear’ to 2 ‘the synesthetic color is definitely more clear’.
Finally, we used a test of type (Projector vs Associator (P/A)) questionnaire (Rouw and Scholte, 2007)). This questionnaire consisted of 12 propositions (6 ‘projector’ items and 6 ‘associator’ items) scored on five-point Likert scales with answer possibilities ranging from 1 ‘strongly disagree’ to 5 ‘strongly agree’. An example proposition is: ‘It is as if the synesthetic color actually originates from the paper on which the letter/number is printed.’ Cronbach's alphas for the 6 ‘projector’ items and 6 ‘associator’ items were .89 and .73, respectively. A P/A value is determined by subtracting the ‘associator’ score from the ‘projector’ score. A positive value represents someone being a ‘projector’. As Eagleman (2012) points out, there exists no good evidence that spatialization among individuals is binary. Rouw and Scholte (2007) found that people scored smoothly along a spectrum rather than in a bimodal distribution. Classifying between ‘projectors’ and ‘associators’, the two proposed forms of synesthesia that describe how you experience your synesthetic percept, should thus be done cautiously.
The software for administering the synesthetic Stroop task was Presentation software® (Version 0.70, www.neurobs.com), since this software is specialized in measuring reaction times very accurately. The Stroop task was done on a Dell computer, running on Windows XP professional. The tests were programmed by S. Beekhuizen (2015), based on original scripts by O. Colizoli (2014).
The online questionnaires used Qualtrics software, Version 56395 of the Qualtrics Research Suite, Copyright © 2014 Qualtrics (Qualtrics, Provo, UT, USA,
http://www.qualtrics.com.).
Data Analysis
Before starting data analysis we had to clean our Stroop data by getting rid of wrong responses and outliers. First, we excluded trials in which the participants gave the wrong answer or failed to answer quickly enough (779 trials, 5.46%). We then looked at where the mistakes during the Stroop task were made: across all participants 637 mistakes (4.46%) were made at incongruent trials in contrast to 142 (1%) at congruent trials. This is in line with what you would expect from participants with real GCS. This left us with only correct Stroop answers and their respective RT’s. Next, we excluded RT’s that fell more than three SD’s from each subject’s mean RT in both the congruent and incongruent Stroop conditions (275 trials (1.93%), 164 congruent (1.15%), 111 incongruent (0.78%), the max number of outliers per person was 12).This then left us with the correct Stroop trials for every participant, falling within each respective interval three SD’s from the mean. On average 32 out of 432 trials (7.4%) were excluded from the analysis for every participant.
We then tested the relevant outcome variables (Stroop effect, CPT, P/A, INT and HSK) for assumptions of normality and homogeneity of variance with Shapiro-Wilk tests and Levene’s test for homoscedasticity.
For the testing of our first hypothesis (Stroop effect) we compared the difference in reaction times (RT’s) between congruent (letter/color pairings that match the synesthetic percept) and incongruent (letter/color pairings that don’t match the synesthetic percept) trials. We did this for all participants to look for an overall Stroop effect, and we also looked at each participant separately. We used dependent t-tests. We expected a significant difference between conditions for all our synesthetes.
For the testing of our second hypothesis (all four measures will correlate positively with each other) we did a correlation analysis. If the data were normally distributed we would test for correlations between the measures with Pearson’s R, using a linear function. Otherwise we could use Spearman’s rank correlation coefficient (rho). Both measures take values between -1 and 1 (1 meaning that one variable perfectly predicts the other). Spearman’s rank correlation is a
nonparametric measure of statistical dependence between two variables (increasing concurrently, but not at the same rate). It uses a monotonic function to investigate this relationship and it can be used for both discrete and continuous variables. In many settings there is only a minimal difference between Pearson’s R and Spearman’s rho correlation coefficients. However,
Spearman’s rho is robust to outliers (unlike Pearson's correlation) and more adjusted to deal with skewed variables. We expected scores on all four measures to correlate positively with each other.
For the testing of our third hypothesis we made between subject comparisons based on difference grouping determined by scores on the P/A questionnaire (e.g. comparing all
‘projectors’ vs all ‘associators’ on the measures obtained in this study). We did this using independent t-tests. We expected ‘projectors’ to score significantly higher on all measures than ‘associators’. All analyses in the study were done using RStudio (version 0.98.1079) and SPSS (version 22).
RESULTS
Confirmatory Analyses
Stroop Effect:
All of our participants showed a positive Stroop effect, ranging from 15 ms to 593 ms (M
= 234 ms). This means all participants showed longer RT’s in the incongruent condition compared to the congruent condition. To check if there was a statistically significant Stroop effect for all our participants we looked at the difference in RT between all congruent (M = 650 ms, SD = 264) vs all incongruent (M = 884 ms, SD = 388) trials. A paired samples t-test showed that indeed there was a significant difference between the means of these two conditions: t(6112) = 41.56, p <.001. Participants showed a synesthetic Stroop effect: their RT’s were significantly longer in the incongruent condition (disagreeing with their synesthesia) compared to the congruent condition (agreeing with their synesthesia). Our first hypothesis is thus confirmed.
Correlation:
All GC synesthetes had CPT scores between 0.43 and 1.94 (M = 0.80). According to Eagleman et al. 2007: ‘a score below 1.0 is ranked as synesthetic. Non-synesthetes asked to use memory or free association typically score in the range of a 2.0. A perfect score of 0.0 would mean that there was no difference in the colors selected on each successive presentation of the same letter.’ Two participants’ CPT scores were unknown. P/A scores ranged between -3.17 and 3.33 (M = -1.15). HSK scores didn’t have a lot of answer possibilities, so its range is also
restricted, from -2 to 2 (the complete scope of its answer possibilities, M = -0.69). INT scores ranged between 1.13 and 4.63 (M = 2.43). To calculate the correlation matrix between our variables of interest we used Spearman’s rho (excluding cases pairwise) since Shapiro-Wilk tests showed the data differed significantly from a normal distribution: CPT (W = 0.86, p = .001),
Stroop (W = 0.93, p = .04), P/A (W = 0.82, p < .001), HSK (W = 0.91, p = 0.01), INT (W = 0.93, p = 0.04). See Table 1 for the correlation analysis results.
Table 1. Correlations and p-values for our five variables under investigation (N=33)
Correlations
Stroop CPT P/A HSK INT
Stroop 1.00 - - - -
CPT 0.17 1.00 - - - P/A -0.11 -0.14 1.00 - - HSK 0.04 0.27 0.14 1.00 - INT 0.06 0.09 0.35* 0.31 1.00 Note: N=31 for correlations involving CPT, due to missing values for two participants
P-values
Stroop CPT P/A HSK INT Stroop - - - - - CPT 0.3711 - - - - P/A 0.5436 0.4487 - - - HSK 0.8165 0.1493 0.4301 - - INT 0.7396 0.6145 0.0435* 0.0843 - *Note: significant at p <.05
As can be seen from the table above only one of these correlations was statistically significant, namely the positive correlation between P/A and INT, so between the subjective experience of projecting the synesthetic color onto graphemes and the extent to which the
synesthetic perception resembles a ‘real’ perception, r(31) = 0.35, p < .05. It is also notable that the correlation between HSK and INT shows a trend towards significance, r(31) = 0.31, p = .08.
Looking at the direction of relationships, not all of them are as expected. The strength of the Stroop effect showed a positive relationship with the CPT consistency scores and a negative relationship with P/A scores, albeit non-significant. This was against our expectation that more consistent synesthetes also suffer more from Stroop interference and ‘project’ their synesthetic percept more. There is no relationship between the Stroop effect and measures of intensity and realism of the synesthetic experience (HSK and INT).
The CPT scores showed a relationship in a negative direction with P/A values. It also showed a positive relationship with HSK, but no relationship with INT. Next, P/A values showed negative non-significant relationships with CPT and the Stroop effect as mentioned above, but it showed a positive trend with HSK. Finally, HSK and INT showed a non-significant positive relationship with each other. Our second hypothesis is thus not confirmed. Our defining criteria for GCS did not correlate in the expected direction (positively).
Projectors vs Associators:
In order to look at the difference between ‘projectors’ and ‘associators’ we first used the P/A questionnaire (Rouw & Scholte, 2007) to differentiate our participants into these two categories. We found that 6 participants were ‘projectors’ (P/A > 0), and 27 participants were ‘associators’ (P/A < 0). This is in line with assumption that being a ‘projector’ is less common since this is a more intense form of synesthesia (Dixon, Smilek & Merikle, 2004).
Since our data were non-normally distributed, we used a non-parametric alternative to the independent t-test, namely the two-sample Wilcoxon test known as the Mann-Whitney test: a nonparametric test of the null hypothesis that two samples come from the same population against an alternative hypothesis that the samples come from different populations, especially where one particular population tends to have larger values than the other.
Using Mann-Whitney tests we tested if there was a meaningful difference between these two groups on the variables of interest. We found no statistically significant differences between ‘projectors’ and ‘associators’ on Stroop (W = 80, p = .53), CPT (W = 36.5, p = .86) and HSK (W = 98.5, p = .21) measurements. We did however find a significant difference between the groups on INT scores, (W = 139, p = .003). This result indicates that there is a difference in the way the two types of GC synesthetes experience their synesthesia. This result is in line with the previously reported result that P/A scores seemed to correlate positively with INT, but negatively with Stroop and CPT.
An important note to make here is that being a ‘projector’ vs an ‘associator’ is not necessarily a dichotomous thing. This is evidenced by the difference in scores on the P/A questionnaire: positive scores implicate someone projects his or her synesthetic percept unto the outside world, whereas negative scores implicate someone only ‘associates’ colors with
letters/numbers in the mind’s eye. The scores ranged from -3.17 to 3.33 for all participants, but even within ‘projectors’ the range of scores had a width of 2.83 points. That is why treating ‘projectors’ and ‘associators’ as two different categories might be the wrong way of looking at this subjective synesthetic characteristic. This is in line with the aforementioned comment by
Eagleman (2012) that ‘there exists no good evidence that spatialization among individuals is binary.’
Exploratory Analyses
Besides our confirmatory analyses of which we had explicit hypotheses we decided to include some other variables to see if they showed interesting patterns in the data. We will go through them one by one. These include Accuracy and Speed of the CPT test, VVIQ-2
(vividness of visual imagery), age and number of different forms of synesthesia experienced. See Table 2 for the correlations (we excluded cases pairwise in calculating Spearman’s rho).
Table 2. Correlations between all our exploratory variables (N=33).
VVIQ-2 ACC Speed Age Number Stroop -0.18 -0.28 0.12 0.17 0.20 CPT 0.24 -0.56** 0.60*** 0.29 0.16 P/A -0.19 0.12 -0.16 0.18 0.07 HSK 0.38 -0.05 0.23 0.31 0.07 INT 0.45* 0.01 0.08 0.28 0.40* VVIQ-2 1.00 - - - - ACC -0.19 1.00 - - - Speed 0.04 -0.25 1.00 - - Age 0.23 -0.35 0.13 1.00 - Number 0.28 -0.37* 0.20 0.28 1.00 Note: not all correlations use N=33 because for all exploratory variables multiple observations were missing *Note: significant at p <.05
**Note: significant at p <.01 ***Note: significant at p <.001
Accuracy and Speed of the CPT test:
Both accuracy on the CPT test and speed on the CPT test correlated with the CPT consistency score in the direction that one would expect, respectively, r(28) = -0.56, p = .001 and r(28) = 0.60, p < .001. Synesthetes with lower consistency scores (which indicate a stronger consistency) had a higher accuracy and a faster reaction time. Furthermore, accuracy and number
of forms of synesthesia negatively correlated with each other, r(28) = -0.37, p = .04 indicating that the more types of synesthesia someone experiences, the less accurate they become.
Vividness of Visual Imagery
The VVIQ-2 (N=22) score measuring how vividly people can imagine objects and scenes correlated positively with the Intensity score, r(20) = 0.45, p = .03. So people had more vivid visual imagery capabilities when the synesthetic experience resembled a ‘true’ experience more.
Age
We looked if there was a correlation between our measures and age. First-person reports suggested that the consistency and intensity of synesthetic experiences tend to wane with age. Since we have a large and diverse range in the age of our synesthetes we were interested if we could replicate these age effects on synesthesia. Correlation analysis showed that age did not correlate significantly with any of our measures.
Number of forms of synesthesia
We also decided to investigate if the number of different forms of synesthesia that someone experiences is related to our variables of interest. This was negatively related to accuracy (see above) and positively correlated to INT, r(29) = 0.40, p = .02. The more forms of synesthesia someone experienced, the more it resembled a ‘real’ perception. It is important to note however, that this measurement is probably not very reliable. This is because the Eagleman test battery was done at home by the participants, and we can’t be sure if they filled it out
honestly and completely. This specific measure of different forms of synesthesia is susceptible to mistakes because often people do not know exactly what constitutes a form of synesthesia or they forget a specific type because it is so familiar to them.
Handedness
Since we had almost no participants who were left-handed (3 participants, 9.1%) we could not investigate if there was a correlation between our measures and handedness.
DISCUSSION
Conclusion
In accordance with our first hypothesis we found a synesthetic Stroop effect for our GC synesthetes. We did not find the expected pattern of correlations however. The only variables that are significantly related in a positive direction are P/A and INT scores. This result makes sense since these two measures both pertain to how much synesthetic experiences resemble and blend with real world properties. Automaticity doesn’t appear to be strongly and significantly related to any of the other defining criteria for GCS. The same goes for consistency. The Stroop effect did show a positive trend with consistency scores, meaning that the bigger the Stroop effect was, the lower the actual consistency, which is not in the expected direction. P/A scores and CPT scores showed a negative trend, which makes sense according to our expectations: the more someone ‘projects’ his or her synesthetic experiences, the higher their color-grapheme consistency is. HSK and INT were also positively related, but this also did not reach significance. This makes sense since these two variables are about how intense the synesthetic experience is. Finally, there appeared to be no differences between ‘projectors’ and ‘associators’ on our
variables of interest, except for INT, measuring the way synesthetes experience their synesthesia. This result stands in contrast with the result of Dixon, Smilek and Merikle (2004) who did find a Stroop difference between these two types of synesthetes.
Since the results do not match our expectations it appears that synesthesia is not as one-dimensional as we thought (or hoped for). It is a more complex interlinked condition of
characteristics then is proposed in the current body of literature (as is already evidenced by the many different forms of inter and intra-modal sensory experiences grouped under the term synesthesia). Our results suggest that GCS should not be defined solely by a consistency criterion since other valid measures of GCS do not seem to significantly correlate with consistency. With these results this study further informs the research field of synesthesia by trying to get a better
idea of how the different defining measures of synesthesia actually relate. Hopefully this marks another step in better understanding synesthesia.
Limitations
There are however a number of methodological issues that could have influenced the results. First of all, The Grapheme CPT measures didn’t look at consistency of the
grapheme/color pairings over time (longitudinally), but at one time point within a session. Both measures of consistency are valid, but because of this the Eagleman Synesthesia Battery cannot say anything about how consistency changes over time. This might prove a problem, because the participants completed the Eagleman Test Battery at various points in time (ranging from 2012 to 2015), but the other measures in the study were all collected in 2015. Therefore, consistency scores might not be completely correct due to this time-gap between the testing of the Grapheme CPT and the other measures.
Moreover, our participants’ CPT scores ranged between 0.43 and 1.94, with most of the scores falling below 1.0. As Eagleman (2007) remarks ‘a score below 1.0 is ranked as synesthetic. Non-synesthetes asked to use memory or free association typically score in the range of a 2.0’. Applying a bigger range of CPT scores to our study might have provided more power to find a significant correlation between the CPT scores and our Stroop effect.
In the current study we also did not take the fact into account that not every synesthete had the same number of graphemes for which they had synesthetic associations. Some
synesthetes might therefore have a higher consistency score because they had fewer synesthetic graphemes because being consistent is easier with fewer observations.
Besides testing consistency there is also an important point to make with respect to the synesthetic Stroop task. In a classical Stroop task one has the option to use non-color words presented in specific colors as a control condition from which to measure Stroop interference and Stroop facilitation. Stroop interference is the effect of slower RT’s because of the
incongruence between the spelled color word and its actual color. Stroop facilitation is the added benefit of presenting a spelled color word in its agreeing color. These two effects can be
separated by using the control condition (neutral words) as a benchmark. There is no such control possible in synesthetic Stroop tasks to measure the exact extent of Stroop facilitation and Stroop interference. When doing synesthetic Stroop tasks one only knows the difference between these two extremes. There is usually no possibility of presenting a ‘neutral’ letter to synesthetes, unless they only have color associations for a specific subset of letters and numbers.
Furthermore, in designing the Stroop task it was decided that when a letter appeared on the screen and subjects had to react to its color as fast as possible, it was first presented in black for about 100 ms before taking on congruent/incongruent colorsto which the subjects had to respond. This was done to prime the synesthetic color experience. However, some participants commented on this saying they found it somehow confusing, even though the graphemes were only presented in black for a very short amount of time. It created conflict for a few participants, where they had to readjust when the grapheme took on its test color.
Lastly, a certain amount of noise is probably inherent in the Stroop data thanks to the way the letters of the alphabet and the colors are linked. To execute the Stroop task, synesthetes had to press four buttons (two on the left side and two on the right side, pressed with middle and index fingers) which had their synesthetic colors assigned to them in a random order. The color-button combination was thus different for everyone, but because of this sometimes the visuo-spatial lay-out of the letters and colors didn’t match; for instance imagine the letter B as being blue. The button for blue is located on the right of the keyboard. This can create interference, since the letter B is at the beginning of the alphabet and the inclination of the synesthete is to move to the left. These are very small effects however and can be partly controlled for by using a button box instead of a keyboard for instance.
Besides the measures of consistency and automaticity used in this study we had
complex as synesthesia is the difficulty in the phrasing of questions pertaining to subjective experience. It is always difficult to find the right words to make sure the participants understand exactly what you mean, so that they can give answers that are interpretable. The way the
questions were phrased for the P/A questionnaire and the INT questionnaire was for instance very similar. They both talked about how much the synesthetic percept resembled real-world percepts and if the subject sometimes confused the two. This might explain the relationship we found between these two measures.
We divided our participants into ‘projectors’ and ‘associators’ based on the results from the P/A questionnaire (Rouw & Scholte, 2007). There appeared to be only 6 ‘projectors’ though, against 27 ‘associators’. This small sample size could be a reason that we did not find a significant difference between these two ‘types’ of synesthetes.
Furthermore, most of the subjects came from or near Amsterdam. Seeing the location of part of the research was at the University of Amsterdam (UvA) this was inevitable since there were also some budget constraints and it was not feasible to compensate everyone for their travel costs. There is no reason to believe however that this overrepresentation of people from or near Amsterdam had a big influence on the results.
. Finally, in selecting synesthetic participants, there was still a reliance on the consistency criterion. This was the first measurement we could easily distribute among a large number of people to see if they would be eligible to participate in the current study. However, incorporating participants with a bigger range of scores on our measures might change the results that we found. Most of the aforementioned issues are, however, about methodological choices rather than methodological problems and they will probably have small effects on the current results. For now we can say that within moderately strong GC synesthetes, our four defining criteria for GCS do not seem to correlate with each other.
This study serves as a first step into investigating how the different defining criteria of GCS actually relate to each other. An important next step would be to replicate the results
obtained in this study with a different sample of GC synesthetes. Since we only tested moderately strong synesthetes (and we did not include controls or people with a wider range of scores) it might be that we did not have enough power to see how our measures actually correlate. If our sample included more diverse participants we might have found a positive correlation. In the future it is thus necessary to include a wider range of scores on the measures of interest and also look at non-synesthete controls. This might reveal a correlational pattern in contrast to the current findings.
The programming scripts for data collection and data analysis used in this study are so designed that they can be used on any other data set and thus be easily compared with the current study. Another important step would be to perform a meta-analysis on studies using different defining measures of GCS with the current results in mind. In conclusion, it is important not to assume that different criteria claiming to measure the same thing are actually positively related to one another and related to the underlying construct that they intent to measure. There are multiple ways to determine GCS and they appear not to be related. This might mean that the measures actually point to different facets of GCS.
REFERENCES
Alvarez, B. & Robertson, L., 2011. The interaction of synesthetic and print color and the role of visual imagery. Journal of Vision, 11(11), 393–393.
Carmichael, D. a. et al., 2015. Validating a standardised test battery for synesthesia: Does ‘The Synesthesia Battery’ reliably detect synesthesia? Consciousness and Cognition, 33, 375–385. Cohen Kadosh, R., & Terhune, D. B. (2012). Redefining synesthesia? British Journal of
Psychology, 103(1), 20-23.
Colizoli, O., Murre, J. M., & Rouw, R. (2014). Defining (trained) grapheme-color synesthesia. Frontiers in human neuroscience, 8.
Colizoli, O., Murre, J. M., & Rouw, R. (2014). Training Synesthetic Letter-color Associations by Reading in Color. Journal of Visualized Experiments, (84), 1–15. Available at:
http://www.jove.com/video/50893/training-synesthetic-letter-color-associations-by-reading-in-color.
Cytowic, R. E. (2002). Synesthesia: A union of the senses (2nd ed.). Cambridge, MA: MIT Press. Dixon, M.J., Smilek, D. & Merikle, P.M., 2004. Not all synaesthetes are created equal: projector
versus associator synaesthetes. Cognitive, affective & behavioral neuroscience, 4(3), 335–343. Eagleman, D.M. et al., 2007. A standardized test battery for the study of synesthesia. Journal of
Neuroscience Methods, 159(1), 139–145.
Eagleman, D. M. (2012). Synesthesia in its protean guises. British Journal of Psychology, 103(1), 16-19.
Jewanski, J., Day, S. A., & Ward, J. (2009). A colorful albino: The first documented case of synesthesia, by Georg Tobias Ludwig Sachs in 1812. Journal of the History of the Neurosciences, 18, 293–303. doi:10.1080/09647040802431946
Lupiáñez, J., & Callejas, A. (2006). Automatic perception and synesthesia: Evidence from colour and photism naming in a stroop-negative priming task.Cortex, 42(2), 204-212.
Mattingley, J.B. et al., 2001. Unconscious priming eliminates automatic binding of colour and alphanumeric form in synesthesia. Nature, 410(6828), pp.580–582.
MacLeod, C. M. (1991). Half a century of research on the Stroop effect: an integrative review. Psychological bulletin, 109(2), 163.
Price, M.C. & Mentzoni, R. a., 2008. Where is January? The month-SNARC effect in sequence-form synaesthetes. Cortex, 44(7), 890–907.
Rich, A. N., Bradshaw, J. L., & Mattingley, J. B. (2005). A systematic, large-scale study of synesthesia: implications for the role of early experience in lexical-colour
associations. Cognition, 98(1), 53-84.
Rouw, R., van Driel, J., Knip, K., & Ridderinkhof, K. R. (2013). Executive functions in synesthesia. Consciousness and cognition, 22(1), 184-202.
Rouw, R. & Scholte, H.S., 2007. Increased structural connectivity in grapheme-color synesthesia. Nature neuroscience, 10(6), pp.792–797.
Simner, J. (2012). Defining synesthesia. British journal of psychology, 103(1), 1-15.
Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643–662.
Van der Veen, F.M. et al., 2014. Grapheme-color synesthesia interferes with color perception in a standard Stroop task. Neuroscience, 258, 246–253.
Ward, J., Li, R., Salih, S., & Sagiv, N. (2007). Varieties of grapheme-colour synaesthesia: a new theory of phenomenological and behavioural differences. Consciousness and cognition, 16(4), 913-931.
