Neuropsychologia 44 (2006) 29182925

Neural correlates of imagined and synaesthetic colours

Anina N. Rich a,b, , Mark A. Williams a,c , Aina Puce d,e , Ari Syngeniotis f , Matthew A. Howard g , Francis McGlone g , Jason B. Mattingley a
b a Cognitive Neuroscience Laboratory, School of Behavioural Science, University of Melbourne, Victoria 3010, Australia Visual Attention Laboratory, Brigham & Womens Hospital, Harvard University, 64 Sidney Street, Cambridge, MA 02139, USA c McGovern Institute, MIT, Cambridge, MA 02139, USA d Centre for Advanced Imaging, Department of Radiology, West Virginia University, USA e Brain Sciences Institute, Swinburne University of Technology, Australia f Brain Research Institute, Austin & Repatriation Medical Centre, Australia g Unilever Research Laboratories, Port Sunlight, UK

Received 23 January 2006; received in revised form 13 June 2006; accepted 15 June 2006 Available online 9 August 2006

Abstract The experience of colour is a core element of human vision. Colours provide important symbolic and contextual information not conveyed by form alone. Moreover, the experience of colour can arise without external stimulation. For many people, visual memories are rich with colour imagery. In the unusual phenomenon of grapheme-colour synaesthesia, achromatic forms such as letters, words and numbers elicit vivid experiences of colour. Few studies, however, have examined the neural correlates of such internally generated colour experiences. We used functional magnetic resonance imaging (fMRI) to compare patterns of cortical activity for the perception of external coloured stimuli and internally generated colours in a group of grapheme-colour synaesthetes and matched non-synaesthetic controls. In a voluntary colour imagery task, both synaesthetes and non-synaesthetes made colour judgements on objects presented as grey scale photographs. In a synaesthetic colour task, we presented letters that elicited synaesthetic colours, and asked participants to perform a localisation task. We assessed the neural activity underpinning these two different forms of colour experience that occur in the absence of chromatic sensory input. In both synaesthetes and non-synaesthetes, voluntary colour imagery activated the colour-selective area, V4, in the right hemisphere. In contrast, the synaesthetic colour task resulted in unique activity for synaesthetes in the left medial lingual gyrus, an area previously implicated in tasks involving colour knowledge. Our data suggest that internally generated colour experiences recruit brain regions specialised for colour perception, with striking differences between voluntary colour imagery and synaesthetically induced colours. 2006 Elsevier Ltd. All rights reserved.
Keywords: Colour imagery; Synaesthesia; fMRI; Visual imagery; V4; Occipitotemporal cortex

1. Introduction Colour perception is a fundamental aspect of human vision (Gegenfurtner & Sharpe, 1999). Although several neuroimaging studies have explored the involvement of early visual areas in visual imagery (e.g., DEsposito et al., 1997; Kosslyn, Thompson, Kim, & Alpert, 1995; Kosslyn, Thompson, & Alpert, 1997; Mellet et al., 2000; Slotnick, Thompson, & Kosslyn, 2005), few investigations have specically examined the neural correlates of colour imagery. Several areas in the ventral

Corresponding author. Tel.: +1 617 768 8815; fax: +1 617 768 8816. E-mail address: rich@search.bwh.harvard.edu (A.N. Rich).

occipital cortex are active when participants view coloured stimuli, or perform colour-related cognitive tasks. These include area(s) in the posterior fusiform gyrus, most often called area V4 (Zeki, 1973; Zeki et al., 1991), and more medial ventral occipital regions that are active particularly during tasks that require retrieval of object-colour knowledge (Chao & Martin, 1999; Martin, Haxby, Lalonde, Wiggs, & Ungerleider, 1995; Price, Moore, Humphreys, Frackowiak, & Friston, 1996). Here, we used fMRI to examine neural activity in these regions in two distinct types of internally generated colour experience. First, we examined colour imagery by having participants imagine the colours of familiar objects. Second, we investigated the involuntary colour experiences that characterise the unusual phenomenon of grapheme-colour synaesthesia.

0028-3932/$ see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2006.06.024

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925

2919

Individuals with grapheme-colour synaesthesia experience vivid sensations of colour when they see or hear particular letters or digits, and these experiences are highly consistent over time (e.g., Baron-Cohen, Wyke, & Binnie, 1987; Rich, Bradshaw, & Mattingley, 2005; Rich & Mattingley, 2002). Crucially, synaesthetic colours arise as an internally generated colour experience without voluntary effort, and are difcult to fully suppress even when there is a desire to do so (Dixon, Smilek, Cudahy, & Merikle, 2000; Mattingley, Rich, Yelland, & Bradshaw, 2001; Mills, Boteler, & Oliver, 1999; Odgaard, Flowers, & Bradman, 1999; Wollen & Ruggiero, 1983). Thus, grapheme-colour synaesthesia provides a unique opportunity to examine the neural correlates of an internally generated but involuntary colour experience. The neuroanatomical loci of colour imagery remain controversial. A number of neuropsychological studies have documented an association between impairments of colour perception and decits in colour imagery in patients with damage to the fusiform gyrus (for review, see Farah, 1988). Other authors, however, argue for a double dissociation between the two functions based on patients who seem to have intact imagery in the context of colour perception decits or vice versa (for review, see Bartolomeo, 2002). Very few studies have explicitly examined the neural correlates of colour imagery in healthy individuals, and the data they have yielded are somewhat ambiguous. Howard et al. (1998) used fMRI to investigate brain activity during a colour imagery task. They presented non-synaesthetes with tasks that required them to imagine colour (e.g., to judge whether a canary is a darker yellow than a banana) or to imagine a set of spatial relations (e.g., to judge the relative angle between the hour and minute hands on a clock face showing 20 to 7). Howard et al. found that although coloured stimuli elicited signicant activity in the posterior fusiform gyrus, corresponding to area V4, colour imagery yielded no signicant activity in the same region. It should be noted, however, that the authors used a whole-brain analysis, which may have reduced the likelihood of detecting a subtle change in V4 activity. In a more recent study, Nunn et al. (2002) failed to nd signicant V4 activation in a group of non-synaesthetes trained on associations between coloured squares and spoken words. In their study, participants were asked to imagine the colour they had been trained to associate with each word, and then to estimate the percentage of colour associations they had correctly imagined. Unfortunately, Nunn et al. did not include an objective measure of their participants performance. Moreover, it is not possible to determine whether they were able to imagine the colours, or simply recalled the colour name associated with the spoken word. The same criticism holds for an earlier study using similar methodology (Goldenberg et al., 1989). The hypothesis that colour imagery recruits regions involved in colour perception therefore needs to be tested with a paradigm in which performance can be measured objectively, and in which participants must imagine colour, rather than merely recalling colour names. There have also been a small number of studies investigating the neural correlates of involuntary colour experiences. Patients with Charles Bonnet syndrome, who experience visual hallu-

cinations following peripheral visual pathology, show activity in the posterior fusiform gyrus for hallucinations that involve colour (Ffytche et al., 1998). For individuals with synaesthesia, who experience colours without chromatic stimulation in the absence of nervous system pathology, the ndings of imaging studies have been mixed. Recent studies indicate that, at least in some individuals, area V4 is active during graphemecolour synaesthesia. Sperling, Prvulovic, Linden, Singer, and Stirn (2006) found increased activity in retinotopically mapped V4 when synaesthetes viewed achromatic letters that induced involuntary colours relative to a condition with letters that did not induce coloured synaesthesia. This pattern was signicant for two of the four synaesthetes when analysed individually. Similarly, Hubbard, Annan, Ramachandran, and Boynton (2005) reported that synaesthetes had greater V4 activity than controls for synaesthetic inducers relative to characters that did not induce synaesthesia. Activity in left V4 has also been documented in a late-blind synaesthete, with more anterior fusiform activation in the right hemisphere when the same participant was asked to imagine colours (Steven, Hansen, & Blakemore, 2006). Finally, Nunn et al. (2002) reported increased V4 activity when synaesthetes heard words that elicited colours compared with non-inducing tones. In contrast, other studies of synaesthesia have failed to nd signicant activation in colour-selective regions of cortex, and have instead documented activations in visual association cortex, including inferior temporal and inferior parietal areas (Paulesu et al., 1995; Weiss, Zilles, & Fink, 2005). These ndings suggest that colour-selective areas of cortex are recruited during internally generated colour experiences, but that there may be differences between colour imagery on the one hand, and synaesthetic experiences on the other. In the present study we used fMRI to characterise brain activation patterns associated with internally generated colour experiences in both synaesthetes and non-synaesthetic controls. We rst identied regions in the ventral occipital cortex that responded selectively to colour by presenting large multicoloured rectangles (Mondrians) in alternating blocks with grey-scale luminance equivalent stimuli (McKeefry & Zeki, 1997). Using the peak activations as centres for regions of interest (ROIs), we then analysed data within these regions from separate colour imagery and synaesthetic colour experiments to examine the neural correlates of these colour experiences in the absence of chromatic stimulation. In one experiment, we had participants perform a colour imagery task, to examine the neural correlates of voluntarily generated colour experiences. Participants viewed greyscale photographs of pairs of objects that have a typical or canonical colour (e.g., a banana and a cob of corn; Fig. 1A). Participants were asked to judge which object, if seen in its natural chromatic form, would be darker in colour (colour imagery task); in a control condition they were asked to indicate which object would be larger in size (size baseline task). The colour imagery task was designed to maximise the likelihood that participants would voluntarily evoke colour images rather than simply drawing on stored verbal colour labels concerning familiar objects. Thus, for example, simply knowing that bananas are yellow should not help in determining whether they are a darker yellow than a

2920

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925

Fig. 1. Example stimuli and fMRI activations for the voluntary colour imagery experiment. (A) Example stimuli for the colour imagery and size (baseline) tasks. Participants judged which object would be darker in colour or larger in size. (B) Activity within the functionally dened ROIs for the colour imagery vs. baseline comparison. Brain slices show regions of signicantly greater activity for control participants (yellow) and synaesthetes (red). For controls signicant activation occurred in the right inferior occipital lobe (xyz: [26 90 18]; p = 0.016); for synaesthetes signicant activation occurred in the right inferior occipital lobe (xyz: [34 88 20]; p < 0.001) and right fusiform gyrus (xyz: [30 94 26]; p = 0.001). Coordinates in this and the subsequent gure have been calculated using the MNI three-dimensional space.

cob of corn. The baseline task, on the other hand, was designed to require a similar judgement of a stimulus attribute that was not related to colour. We used identical pairs of pictures in the two tasks to ensure that any changes in brain activity could not be due to low-level luminance, intensity, contrast, or featural differences between stimuli. In a second experiment, synaesthesia-inducing letters were presented either in the matching or congruent colour for each synaesthete, or in shades of grey (achromatic) equated for luminance. These conditions were compared with a baseline condition of greyscale placeholders (Fig. 2A). A group of sex-, age- and handedness-matched non-synaesthetes viewed the identical stimuli. Participants were required to detect the brief disappearance of one stimulus in the set of four presented in each display in a localisation task.

We hypothesized that if regions of the brain specialised for colour perception are also involved in colour imagery, both synaesthetes and controls should show signicant activation within the ROIs dened by the colour localiser. Similarly, if synaesthetic colours recruit areas involved in colour perception, signicant activation should be present within the same ROIs for synaesthetes, but not for controls, during presentation of achromatic letters.
2. Methods 2.1. Participants
Seven grapheme-colour synaesthetes (6 female; 5 right-handed; mean age = 28 years, S.D. = 7.1, range: 1837 years), and seven non-synaesthetic controls matched for age, sex and handedness (mean age = 27 years; S.D. = 7.4;

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925 range: 1736 years) participated after giving written informed consent according guidelines prescribed by the University of Melbourne Human Research Ethics Committee. The mean ages of the two groups did not differ significantly, t(12) = 0.22, p > 0.05. Data from one of the synaesthetes were lost from the colour perception and colour imagery tasks due to a scanner hard disk error, leaving six synaesthetes in these conditions. All participants were classied as normal by neuroradiological review; all had normal or correctedto-normal visual acuity, and normal colour vision as assessed by the Ishihara colour plates (Ishihara, 1975). All of the synaesthetes reported their synaesthetic colours to appear in the minds eye, although more detailed descriptions varied from the letters appearing in colour to a coloured overlay. Two synaesthetes classied themselves as having minds eye colours while also describing the colour as appearing on the page or on a dark screen.

2921

Participants viewed the screen through a mirror mounted on the quadrature headcoil. Optic-bre button boxes were used to record participant behavioural responses (reaction time and accuracy). To synchronise timing between stimulus presentation and image acquisition, a trigger pulse from the MR scanner initiated the start of each block within an imaging run.

2.3. Stimuli and procedure

Initially, participants completed two blocks of each task outside the scanner to familiarize themselves with the tasks. They then performed each task concurrently with fMRI acquisition. 2.3.1. Colour perception localiser Both groups of participants viewed alternating 21-s blocks of coloured and greyscale luminance-matched Mondrians (visual angle = 19.5 19.5 ). Each stimulus was presented for 1000 ms, with a blank grey interstimulus interval (ISI) of 500 ms. There were eight alternating blocks within the run (four coloured; four greyscale), with a total run time of 168 s. 2.3.2. Colour imagery task Participants viewed greyscale photographs of pairs of familiar objects (Fig. 1A), and indicated via an index-nger button on each hand which object

2.2. Apparatus
During initial behavioural testing outside the scanner, visual stimuli were presented centrally on a 15-in. CRT monitor (ViewSonic Graphics Series G771, 800 600 pixel screen resolution, 60 Hz refresh rate). During fMRI acquisition the same images were back-projected via a Sony SVGA LCD data projector (VPL-SC1) onto an opaque screen positioned at the end of the scanner gurney.

Fig. 2. Stimuli, behavioural and fMRI data for the synaesthetic colour experiment. (A) Example displays from the chromatic, achromatic and baseline conditions. Participants identied the location of a 100 ms offset of one of the four stimuli. (B) Behavioural data collected during image acquisition. Mean accuracy (% correct; +1 S.D.) for the localisation task in the chromatic (black bars), achromatic (white bars) and baseline (grey bars) conditions, plotted separately for synaesthetes and controls. Note that chance is approximately 25% in this task (four locations, plus occasional catch trials in which no response was required). (C) Brain slices showing activity within the functionally dened ROIs for chromatic vs. baseline (blue) and achromatic vs. baseline (red) for synaesthetes. For the chromatic vs. baseline contrast, there was signicant activation in the left medial ventral lingual gyrus (xyz: [4 80 6], p < 0.001; [6 84 14], p = 0.001), and in a more lateral ventral occipital region (xyz: [20 74 16], p = 0.011). For the achromatic vs. baseline contrast, there was again signicant activity in the left medial lingual gyrus (xyz: [4 86 12], p = 0.008; [2 82 6], p = 0.015). (D) Brain slices showing activity within the functionally dened ROIs obtained for the chromatic vs. baseline (yellow) contrast for controls. There was signicant activation in the right posterior fusiform gyrus (xyz: [30 86 26]; p = 0.015). Additional activity (not shown) was found in the parahippocampal gyrus (xyz: [18 38 12], p = 0.009; [16 34 4]; p = 0.015). There were no signicant voxels within the ROIs for controls in the achromatic vs. baseline contrast.

2922

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925

Fig. 2. (Continued ).

would be darker in colour (colour imagery) or larger in size (size baseline). There were 15 different objects, each of which was presented at least once in each condition (colour imagery and size baseline). The objects were positioned such that approximately the same number of responses would be made with each hand. On average, the objects subtended a visual angle of 7.2 7.6 . Each stimulus pair was presented for 6000 ms, during which participants made their response. This was followed by a blank ISI of 1000 ms. Three object pairs were presented in each block; at the commencement of each block a display appeared for 5000 ms (1000 ms ISI) instructing participants to perform either the colour or the size judgement task. The task alternated across blocks. The order of stimuli was counterbalanced such that the same stimulus pairs were never presented in immediately consecutive blocks. There were 12 blocks within the run, six of each condition, with a total run time of 324 s. Behavioural responses from the two imagery tasks were compared with those obtained from an independent group of seven non-synaesthetic participants who rated the identical achromatic object pairs in a separate pilot study. 2.3.3. Synaesthetic colour task During an initial screening experiment, synaesthetes selected display colours that best matched their synaesthetic colours for four letter stimuli, using a standard computer-generated palette (Mattingley et al., 2001). These letters were

used to construct a unique set of stimuli for each synaesthete. Non-synaesthetic participants viewed the identical stimuli to the synaesthete with whom they were matched, to control for any extraneous effects of particular letter-colour pairings. Participants viewed interleaved blocks of letters (which induced vivid colour experiences for the synaesthetes and baseline stimuli). The letters were coloured congruently for each synaesthete (chromatic condition), or were greyscale versions that were luminance-matched (achromatic condition). The baseline condition consisted of greyscale, luminance-equivalent rectangular placeholders that did not induce synaesthesia. Four stimuli were presented on each trial, arranged symmetrically around a xation point (Fig. 2A). Each of the four coloured letters, or their greyscale equivalents, appeared with equal probability in each location. On average, each stimulus subtended a visual angle of 5.7 6.6 . Participants were required to indicate as quickly and accurately as possible which item within each display briey (100 ms) disappeared. In 70% of blocks, there was a catch trial in which no stimulus disappeared. Trial duration was 2500 ms, with an intertrial interval of 500 ms. The onset of target events varied unpredictably between 400 and 1600 ms (see Fig. 2A). The conditions alternated across blocks in the order baselinechromaticachromatic. There were 18 blocks within the run, six of each condition, with a total run time of 378 s. Thus, each block consisted of seven trials of the same condition, giving a total of 42 trials per condition (126 trials in total).

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925 2.3.4. Imaging data acquisition and analysis All images were acquired on a 3T GE Signa MRI scanner. Functional MRI data were acquired using an oblique axial gradient-echo echo-planar (EPI) sequence (TR = 3 s, TE = 40 ms, 128 128 matrix, FOV = 250 mm, slice thickness = 4 mm, gap = l mm, 22 slices; in-plane resolution = 1.95 mm 1.95 mm). We ensured the ventral and lateral occipitotemporal cortices were completely sampled in our whole-brain imaging protocol. As a result, acquisition over the dorsal frontal areas was incomplete for some participants. The rst four volumes of each functional MRI run were discarded automatically during data acquisition to allow transverse magnetization to reach steady state. A T1weighted sequence (spin-echo, 256 256 matrix, 22 slices of 4 mm thick with 1 mm gap; TE = 14 ms, TR = 500 ms, NEX = 1, FOV = 250 mm; in-plane resolution = 0.98 mm 0.98 mm) was performed to obtain anatomical detail in a brain volume with orientation identical to the functional MRI study. Pre-processing and data analysis were performed using SPM2 (Wellcome Department of Imaging Neuroscience; http://www.l.ion.ucl.ac.uk/spm/ spm2.html). To correct for head motion, functional MRI volumes were realigned within scanning runs to the rst image using a set of six rigid body transformations determined for each image. Each functional MRI run was spatially normalized to the EPI-template supplied with SPM2, beginning with a local optimisation of the 12-parameters of an afne transformation. These images were then smoothed with an 8 mm Gaussian lter. The BOLD signal was analysed within each run using a high-pass frequency lter (cut-off 144 s). Correlations between scans and epochs were modelled by a standard haemodynamic response function at each voxel. Parameter estimates were obtained for each condition and participant to allow second-level random effects analysis of between-participant variability. The regions of signicant activity for these analyses are illustrated using overlays on standard MNI space templates (Wellcome Department of Imaging Neuroscience; http://www.l.ion.ucl.ac.uk/spm/spm2.html). Peak activations from the localiser scan were used to dene the centre of each 10 mm ROI using WFU Pickatlas (Maldjian, Laurienti, Burdette, & Kraft, 2003) within SPM2. These regions were then interrogated for the experimental scans, reducing the number of multiple comparisons. Note that within the ROIs, we performed small-volume analyses, rather than calculating an average across all voxels within each region.

2923

letter conditions (compared to chance of 25%), demonstrating they were effectively attending to the display. 3.2. Functional MRI data 3.2.1. Colour perception localiser We localised regions of ventral occipital cortex that were selectively responsive to the coloured Mondrians relative to the matched greyscale stimuli using random-effects (second-level) analyses, restricted to the fusiform and lingual gyri within the occipital lobes (selected using WFU Pickatlas; Maldjian et al., 2003). As we planned to interrogate all potentially active voxels for subsequent experimental runs, we set a liberal threshold of p < 0.05, and included all participants as a single group for this initial analysis. Presentation of coloured stimuli resulted in signicant activation in the left and right lingual gyri, and in the right posterior fusiform gyrus. We generated 10 mm spheres around the peak activations within extrastriate cortex (left peaksMNI coordinates xyz: [12 90 22]; [8 92 12]; [20 84 16]; [4 66 4]; [2 74 0]; right peaksxyz: [16 42 4]; [28 88 26]), forming seven colour-perception ROIs for all subsequent analyses. Although there were no peak activations in the left fusiform gyrus, much of the range described by McKeefry and Zeki (1997) for left V4 was included in the ROI surrounding the peak [20 84 16]. The ndings of the chromatic condition of the synaesthetic colour task (see below) conrm that this ROI includes left V4. 3.2.2. Colour imagery task Colour imagery data were analysed separately for synaesthetes and controls. Interrogation of voxels within the colourperception ROIs for the colour imagery versus baseline comparison revealed signicant activation in right posterior ventral occipital cortex, both in synaesthetes and controls (Fig. 1B). The location of activity during voluntary colour imagery in both groups was consistent with that reported for the human homologue of monkey area V4. The average group activity in right V4 reported by McKeefry and Zeki (1997) extended in x from 24 to 36, in y from 71 to 83, and in z from 18 to 33 (converted from the original coordinates to MNI coordinates for comparability),1 with the most individual variability in the location of V4 in the anteriorposterior (y) plane. In our study, the coordinates for the right ventral activity peaks in the synaesthete group were xyz: [34 88 20] and [30 94 26], and in the controls were xyz: [26 90 18]. There were no other signicant activations within the ROIs. We recognize that there is debate about the presence of other colour-selective regions in the posterior fusiform gyrus (e.g., V8; Hadjikhani, Liu, Dale, Cavanagh, & Tootell, 1998). Hence, we localised our colourselective ROIs functionally rather than relying on coordinates derived from previous investigations. For convenience, and to
1

3. Results 3.1. Behavioural data 3.1.1. Colour imagery task Behavioural data collected during scanning conrmed that participants performed the task as instructed. Responses of both the synaesthetes and controls matched those of the pilot group in more than 70% of trials (synaesthetes: colour imagery task M = 74%, S.D. = 6%, size baseline task M = 79%, S.D. = 3%; controls: colour imagery task M = 72%, S.D. = 7%, size baseline task M = 75%, S.D. = 7%). 3.1.2. Synaesthetic colour task Behavioural data collected during scanning indicated that both groups performed the task as instructed (Fig. 2B). Controls performed similarly across conditions (chromatic, achromatic and baseline), whereas synaesthetes were more accurate in the baseline condition than in either of the letter conditions (group by condition interaction, F(2,24) = 4.70, p < 0.05; controls: p > 0.05 for all pairwise comparisons; synaesthetes: chromatic versus baseline, p < 0.05; achromatic versus baseline, p < 0.05; chromatic versus achromatic, p > 05). This difference was unexpected, and was not observed in the behavioural data collected prior to scanning. Importantly, however, during scanning synaesthetes still performed at a rate of 70% correct in the

Original coordinates based on the ICBM/NIH 10 template from McKeefry and Zeki (1997) for average group activity in V4. Right V4: x, 24 to 36; y, 70 to 82; z, 12 to 24. Left V4: x, 20 to 36; .y, 66 to 84; z, 8 to 18.

2924

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925

maintain consistency with the literature, we use the term V4 to denote these functionally dened ROIs. 3.2.3. Synaesthetic colour task For synaesthetes, interrogation of voxels within the colourperception ROIs for the comparison of greyscale letters versus baseline revealed signicant activity in the left medial lingual gyrus (xyz: [4 86 12]; Fig. 2C). By contrast, there were no signicant activations within the same ROIs for nonsynaesthetic controls. For the comparison between chromatic letters and the greyscale baseline, synaesthetes showed significant activations within the left medial occipital cortex. The synaesthetes also had signicant activity in a more lateral region of the occipital cortex corresponding to left V4 (xyz: [20 74 16]), demonstrating an additional response to the presence of colour in the display. For the achromatic letters versus baseline comparison, there was signicant activity in the left medial occipital cortex, and this overlapped with the activation from the chromatic letters versus baseline comparison (xyz: [4 80 6]; [6 84 14]; Fig. 2C). No other brain regions within the ROIs showed signicant activity. Controls showed a significant response in right V4 only when viewing coloured letters (xyz: [30 86 26]; Fig. 2D). Unlike the synaesthetes, the controls showed no activity in the medial lingual gyrus in either the chromatic or achromatic conditions. 4. Discussion The aim of this study was to explore the neural correlates of two forms of internally generated colour experience, focusing in particular on regions within the ventral occipital cortex known to respond selectively to coloured stimuli. The results demonstrate that both colour imagery and synaesthetic colours recruit ventral occipital regions, but there were differences in the loci of activation between the two types of colour experience. Voluntary imagery activated right V4 in both synaesthetes and controls, whereas synaesthetic colours were associated with activity in a more medial region of the left lingual gyrus in our synaesthetic subjects only. Importantly, this latter activity was not present in non-synaesthetic controls, who performed the same task with identical stimuli. This difference between synaesthetes and controls suggests that the left lingual gyrus activity observed in synaesthetes cannot be attributed to letter recognition alone. When coloured letters were presented, V4 activity was evident for both groups, though on the left in the synaesthetes and on the right in the controls. Although it is tempting to speculate that this laterality difference might reect anomalous colour processing in the synaesthetes (e.g., see Nunn et al., 2002), it is important to note that the laterality and position of V4 activation varies considerably even among non-synaesthetic individuals (for discussion, see McKeefry and Zeki, 1997). Indeed, in the Nunn et al. (2002) study, it was only right V4 that was activated by chromatic stimuli in synaesthetes, the opposite pattern from the current results. Our results differ from those of a previous study by Howard et al. (1998), which failed to demonstrate V4 activation during colour imagery in non-synaesthetes. This discrepancy might

be due to the increased power to detect subtle effects afforded by our region of interest approach (cf. the whole-brain analysis used by Howard et al.). In addition, Howard et al. used a baseline imagery task judging the angle between hands on a clockface which involved different stimuli and task demands from their colour imagery task. To overcome these problems we used identical stimuli in our two imagery conditions (colour and size), and attempted to match the task demands as closely as possible. Thus, the brain activations we found cannot be attributed to low-level differences between displays. This is particularly important given evidence that V4 is sensitive to changes in luminance (Motter, 1994; Reynolds, Pasternak, & Desimone, 2000). For the synaesthetic participants, we observed robust activity in the left medial lingual gyrus for both conditions in which synaesthesia-inducing characters were present. The left lingual gyrus has been implicated in a number of colour-related tasks in non-synaesthetic participants, including naming of display colours (relative to viewing colour; Price et al., 1996), generating colour names for achromatic objects (Chao and Martin, 1999), and other tasks that require retrieval of colour knowledge (Martin et al., 1995). Thus, the involvement of this region in synaesthetic colour experience is consistent with the notion that the left lingual gyrus is recruited by tasks requiring access to colour representations. In contrast to some previous investigations (Hubbard et al., 2005; Nunn et al., 2002; Sperling et al., 2006), we did not observe signicant activity in area V4 during synaesthetic colour experiences. Other authors have proposed individual variability in the extent to which V4 is involved in synaesthesia (Hubbard et al., 2005), which may account for this null effect. Alternatively, it may be due to variability in the locus of V4 across individuals, although the signicant activity we found during the colour imagery and chromatic letters conditions would argue against this latter explanation. We are interested in activation that is unique to synaesthetes and consistent across the group, hence, we concentrated on the robust activity we observed in the left medial lingual gyrus, which clearly suggests a role for this region in synaesthetic colour experiences. Interestingly, a PET study by Paulesu et al. (1995) found a signicant deactivation in the left lingual gyrus when synaesthetes listened to words which elicited synaesthesia relative to non-inducing tones. At this point, we can only speculate on the exact role the lingual gyrus plays in synaesthesia. It seems probable that synaesthesia involves retrieval of colour knowledge and semantic information related to colour, suggesting that the lingual gyrus plays a similar role in synaesthetes and non-synaesthetes. It remains to be determined whether retrieval of such information is the endpoint, or simply a step in the process of generating synaesthetic colours. In either case, it will be important for theories of synaesthesia to take into account the contribution of stored colour knowledge. In summary, the results of the colour imagery experiment demonstrate that voluntary colour imagery engages regions of the extrastriate cortex that respond selectively to coloured stimuli. Our data provide support for the suggestion that imagined stimuli utilise some of the regions of the brain specialised for perception of sensory information (Ganis, Thompson, &

A.N. Rich et al. / Neuropsychologia 44 (2006) 29182925

2925

Kosslyn, 2004; Kosslyn, Thompson, Kim, et al., 1995; Kosslyn, Thompson, et al., 1997). Further, the data reveal striking differences between synaesthetes and controls that implicate the left lingual gyrus in synaesthetic processing, raising the possibility that synaesthesia might rely on retrieval of higher-level conceptual information about colour. Acknowledgements This study was funded by the School of Behavioural Science, University of Melbourne, Australia, and Unilever Research Laboratories, Port Sunlight, UK. ANR is supported by a grant from the National Health & Medical Research Council, Australia. References
Baron-Cohen, S., Wyke, M. A., & Binnie, C. (1987). Hearing words and seeing colours: An experimental investigation of a case of synaesthesia. Perception, 16, 761767. Bartolomeo, P. (2002). The relationship between visual perception and visual mental imagery: A reappraisal of the neuropsychological evidence. Cortex, 35(3), 357378. Chao, L. L., & Martin, A. (1999). Cortical regions associated with perceiving, naming, and knowing about colors. Journal of Cognitive Neuroscience, 77(1), 2535. DEsposito, M., Detre, J. A., Aguirre, G. K., Stallcup, M., Alsop, D. C., Tippet, L. J., et al. (1997). A functional MRI study of mental image generation. Neuropsychologia, 35(5), 725730. Dixon, M. J., Smilek, D., Cudahy, C., & Merikle, P. M. (2000). Five plus two equals yellow. Nature, 406, 365. Farah, M. (1988). Is visual imagery really visual? Overlooked evidence from neuropsychology. Psychological Review, 95(3), 307317. Ffytche, D. H., Howard, R. J., Brammer, M. J., David, A., Woodruff, P., & Williams, S. (1998). The anatomy of conscious vision: An fMRI study of visual hallucinations. Nature Neuroscience, 7(8), 738742. Ganis, G., Thompson, W. L., & Kosslyn, S. M. (2004). Brain areas underlying visual mental imagery and visual perception: An fMRI study. Cognitive Brain Research. Gegenfurtner, K. R., & Sharpe, L. T. (Eds.). (1999). Color vision: From genes to perception. Cambridge, UK: Cambridge University Press. Goldenberg, G., Podreka, I., Uhl, F., Steiner, M., Willmes, K., & Deecke, L. (1989). Cerebral correlates of imagining colours, faces and a map. 1. SPECT of regional cerebral blood ow. Neuropsychologia, 27(11/12), 13151328. Hadjikhani, N., Liu, A. K., Dale, A. M., Cavanagh, P., & Tootell, R. B. H. (1998). Retinotopy and color sensitivity in human visual cortical area V8. Nature Neuroscience, 7(3), 235241. Howard, R. J., Ffytche, D. H., Barnes, J., McKeefry, D., Ha, Y., Woodruff, P. W., et al. (1998). The functional anatomy of imagining and perceiving colour. Neuroreport, 9, 10191023. Hubbard, E. M., Annan, A. C., Ramachandran, V. S., & Boynton, G. M. (2005). Individual differences among grapheme-colour synesthetes: Brainbehavior correlations. Neuron, 45, 975985. Ishihara, S. (1975). Ishiharas tests for colour-blindness. Tokyo: Kanehara Shuppan. Kosslyn, S. M., Thompson, W. L., & Alpert, N. M. (1997). Neural systems shared by visual imagery and visual perception: A positron emission tomography study. Neuroimage, 6, 320334. Kosslyn, S. M., Thompson, W. L., Kim, I. J., & Alpert, N. M. (1995). Topographical representations of mental images in primary visual cortex. Nature, 378, 496498.

Maldjian, J. A., Laurienti, P. J., Burdette, J. B., & Kraft, R. A. (2003). An automated method for neuroanatomical and cytoarchitectonic atlas-based interrogation of fMRI datasets. Neuroimage, 19, 12331239. Martin, A., Haxby, J. V., Lalonde, F. M., Wiggs, C. L., & Ungerleider, L. G. (1995). Discrete cortical regions associated with knowledge of color and knowledge of action. Science, 270, 102105. Mattingley, J. B., Rich, A. N., Yelland, G., & Bradshaw, J. L. (2001). Unconscious priming eliminates automatic binding of colour and alphanumeric form in synaesthesia. Nature, 410, 580582. McKeefry, D. J., & Zeki, S. (1997). The position and topography of the human colour centre as revealed by functional magnetic resonance imaging. Brain, 120, 22292242. Mellet, E., Tzourio-Mazoyer, N., Bricogne, S., Mazoyer, B., Kosslyn, S. M., & Denis, M. (2000). Functional anatomy of high-resolution visual mental imagery. Journal of Cognitive Neuroscience, 72(1), 98115. Mills, C. B., Boteler, E. H., & Oliver, G. K. (1999). Digit synaesthesia: A case study using a Stroop-type test. Cognitive Neuropsychology, 16(2), 181 191. Motter, B. C. (1994). Neural correlates of attentive selection for color or luminance in extrastriate area V4. The Journal of Neuroscience, 14(4), 21762189. Nunn, J. A., Gregory, L. J., Brammer, M., Williams, S. C. R., Parslow, D. M., Morgan, M. J., et al. (2002). Functional magnetic resonance imaging of synesthesia: Activation of V4/V8 by spoken words. Nature Neuroscience, 5(4), 371375. Odgaard, E. C., Flowers, J. H., & Bradman, H. L. (1999). An investigation of the cognitive and perceptual dynamics of a colour-digit synaesthete. Perception, 28, 651664. Paulesu, E., Harrison, J., Baron-Cohen, S., Watson, J. D. G., Goldstein, L., Heather, J., et al. (1995). The physiology of coloured hearing: A PET activation study of colour-word synaesthesia. Brain, 118, 661 676. Price, C. J., Moore, C. J., Humphreys, G. W., Frackowiak, R. S. J., & Friston, K. J. (1996). The neural regions sustaining object recognition and naming. Proceedings of the Royal Society of London B, 263, 1501 1507. Reynolds, J. H., Pasternak, T., & Desimone, R. (2000). Attention increases sensitivity of V4 neurons. Neuron, 26, 703714. Rich, A. N., Bradshaw, J. L., & Mattingley, J. B. (2005). A systematic, largescale study of synaesthesia: Implications for the role of early experience in lexical-colour associations. Cognition, 98(1), 5384. Rich, A. N., & Mattingley, J. B. (2002). Anomalous perception in synaesthesia: A cognitive neuroscience perspective. Nature Reviews Neuroscience, 3, 4352. Slotnick, S. D., Thompson, W. L., & Kosslyn, S. M. (2005). Visual mental imagery induces retinotopically organized activation of early visual areas. Cerebral Cortex, 15, 15701583. Sperling, J. M., Prvulovic, D., Linden, D. E. J., Singer, W., & Stirn, A. (2006). Neuronal correlates of colour-graphemic synaesthesia: A fMRI study. Cortex, 42(2), 295303. Steven, M. S., Hansen, P. C., & Blakemore, C. (2006). Activation of colorselective areas of the visual cortex in a blind synesthete. Cortex, 42(2), 304308. Weiss, P. H., Zilles, K., & Fink, G. R. (2005). When visual perception causes feeling: Enhanced cross-modal processing in grapheme-color synesthesia. Neuroimage, 28(4), 859868. Wollen, K. A., & Ruggiero, F. T. (1983). Colored-letter synesthesia. Journal of Mental Imagery, 7(2), 8386. Zeki, S. (1973). Colour coding in rhesus monkey prestriate cortex. Brain Research, 53, 422427. Zeki, S., Watson, J. D. G., Luecke, C. J., Friston, K. J., Kennard, C., & Frackowiak, R. S. J. (1991). A direct demonstration of functional specialization in human visual cortex. Journal of Neuroscience, 11, 641 649.