Task difficulty in a simultaneous face matching task modulates activity in face fusiform area

Cognitive Brain Research 25 (2005) 701 – 710 www.elsevier.com/locate/cogbrainres

Research Report

Task difficulty in a simultaneous face matching task modulates activity in face fusiform area A.L.W. Bokdea,*, W. Dong a, C. Bornb, G. Leinsingerb, T. Meindlb, S.J. Teipela, M. Reiserb, H. Hampela a

Dementia and Neuroimaging Research Section, Alzheimer Memorial Center and Geriatric Psychiatry Branch, Department of Psychiatry, Ludwig-Maximilian University, Nussbaumstr. 7, 80336 Munich, Germany b Institute of Diagnostic Radiology, Ludwig-Maximilian University, Munich, Germany Accepted 1 September 2005 Available online 1 December 2005

Abstract The level of difficulty of a task can alter the neural network that activates for performance of the task. Previous studies have shown increased activation with task difficulty in the frontal lobes while the effects in the extrastriate visual areas have been unclear. We hypothesized that the face fusiform area (FFA), an area specialized for face processing, would increase activation as task difficulty increased in a face matching task. The difficulty level was increased by degrading the quality of the images. The degradation levels were 10%, 20%, 40% and 60%. Based on the correct response rate, the data were divided into a baseline level (composed of non-degraded and 10% degraded images) and a difficult level (composed of the 20%, 40% and 60% degraded images). Brain activation was measured using functional magnetic resonance imaging. The baseline face matching task activated a wide network of regions that included bilaterally the occipital, temporal and parietal lobes and the right frontal lobe. A novel behavioral finding was that task difficulty did not linearly increase with image degradation. The novel brain imaging finding was that the FFA is modulated by task difficulty and performance in the task was linearly correlated to activation in FFA. In addition, we found that activation in the dorsolateral prefrontal cortex (DLPFC) had increased activation as task difficulty increased. When adding the response time as a covariate, the differences in the DLPFC did not remain statistically significant. Increased task difficulty also led to a decrease in activation of visual areas in the extrastriate cortex. Task difficulty increased activation in the FFA to enhance the face processing and suppressed activation in visual extrastriate areas that processed low level properties of the stimuli. Task difficulty led to enhanced response in the FFA and suppressed response in other visual areas. D 2005 Elsevier B.V. All rights reserved. Theme: Neural basis of behavior Topic: Cognition Keywords: Functional magnetic resonance imaging; Face processing; Object matching

1. Introduction The human brain adapts to the demands placed upon it. Increasing the amount of effort to perform a perceptual task can affect how information in the visual system is processed. Task difficulty can be increased through stimulus degradation, which decreases the perceptual * Corresponding author. Fax: +49 89 5160 5808. E-mail address: [email protected] (A.L.W. Bokde). 0926-6410/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.cogbrainres.2005.09.016

discriminability and enhances decision. One specialized perceptual area is the ‘‘face fusiform area’’ (FFA) defined as a region selectively activated in response to face stimuli (among them [15,43,45,77,80]). Previous studies have shown that the FFA was modulated by selective attention [60,80], and by working memory load [19]. Single unit recording in primates have documented the existence of face-selective and object selective neurons in the inferior temporal lobe and superior temporal sulcus [4,5,67,71,82]. In humans, the homologue to primate face

702

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

selective region is the face selective region in the fusiform gyrus. Given that the FFA can be modulated by attention and working memory load, one would expect that task difficulty would also modulate the FFA, with greater activation as task difficulty increased. Previous studies have found that activation in the early sensory areas decreased [30] or increased [39] as task difficulty was increased. A possible factor that may explain the differences was the decrease in the number of trials as task difficulty increased [30] or the loudness of the noise in the auditory stimuli increased as task difficulty increased [39]. Thus there was a confounding variable in investigating the effects of task difficulty in the sensory processing areas. We wished to investigate this issue further, in particular, the behavior of the FFA. We hypothesized that the FFA would increase activation as task difficulty increased. In addition, we thought it unlikely that activation in other visual areas would be suppressed. The effects of difficulty level on the FFA was investigated in a simultaneous perception-matching task utilizing the blood oxygen dependent level (BOLD) signal measured using functional magnetic resonance imaging (FMRI) as a measure of brain activation. In addition to our prediction of the FFA, we expected that the right DLPFC and ACC would be modulated by task difficulty as was shown by previous studies [30,40]. We examined the activation magnitude using a region of interest (ROI) analysis that included four ROIs in the right hemisphere: the DLPFC, the ACC, the FFA and the anterior fusiform gyrus. The ROI in the anterior fusiform gyrus was to demonstrate that the effects of task difficulty in the FFA, DLPFC and ACC were not due to differences in arousal. Task difficulty was modulated by varying the image quality of the faces presented from no degradation to 60% of the pixels in the face image replaced by random valued pixels. Based on the accuracy of the responses for each run, we divided the different runs into a baseline level and a difficult level. The number of trials and luminance of the images were kept constant across the different runs. In addition, we replicated previous findings using a voxel-based analysis approach.

2.2. Stimuli and tasks Two faces were presented simultaneously and participants were asked to decide in each trial if a pair of faces was identical or not. If identical, the subject responded by pressing a button held in the right hand using the thumb. No response was required if the faces were dissimilar. Each trial in the task, as illustrated in Fig. 1, had 2 squares in which two identical or dissimilar faces were placed. The faces were grey scale stimuli where only the face was visible. Each trial was 2.8 s long with an interval between pairs of faces of 0.318 s. There were 8 trials per block and there were 3 blocks of the task in each run. There were 4 blocks of the control task in each run. At the beginning of each block, there was a 7.2-s task instruction. There were 5 face matching tasks which were: no degradation of the images, 10%, 20%, 40% and 60% degradation. There was one run for each of the face degradation levels. The different conditions were counter-balanced across subjects. Eighty percent of the trials had identical pair of faces. The 10% degradation was measured on 11 from 14 subjects (with 3 subjects the acquisition of the response was not recorded due to equipment malfunction) and the 60% condition was measured on 10 from 14 subjects (data from 4 subjects not measured by chance). The faces were obtained from the

2. Methods 2.1. Subjects There were 14 healthy right-handed subjects (6 men/ 8 women) with an average age (standard deviation) of 26.8 (4.9) years. All subjects gave written informed consent to participate in the study after the study was explained to them. The subjects did not have a history of neurological or psychiatric illness. The study was performed in accordance with the Declaration of Helsinki and the Ethics Committee of the Faculty of Medicine of Ludwig Maximilian University approved the study. All subjects had normal or corrected to normal vision.

Fig. 1. Illustration of the cognitive task with a stimulus example with no degradation and one with 40% degradation. Random pixels were selected from each image, and converted to grey level such that the luminance did not change from the original image.

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

Max Planck Institute for Biological Cybernetics database [8]. The images were degraded by replacing pixels in the image by random grey levels. The luminance of the images was not changed. In the control task for each scan, the subject had to press the button every time a control image appeared (8 times per block). In the conditions were the faces were degraded by 10%, 20%, 40% or 60% the control images were also degraded to the same level. There were 4 blocks of the control task, which were the same length as the face matching task. The parameters for the presentation of the images were identical as with the face matching task. The number of responses during the control task blocks were constant across all runs. The accuracy and reaction time of the responses during performance of the task were recorded for later analysis. The subjects did not practice the task before the scan. 2.3. Scanning The imaging sequence was an interleaved T2* weighted echo-planar imaging (EPI) sequence with 28 axial slices (4 mm slice thickness and slice gap = 1 mm, repetition time (TR) = 3.60 s, echo time (TE) = 60 ms, flip angle = 90 -, field of view (FOV) = 240 mm., matrix = 64  64) and 69 volumes acquired per scan on a 1.5 T Siemens Magneton Vision scanner (Erlangen, Germany). For anatomical reference in each subject, a T1 weighted sequence with 28 slices was acquired in the same orientation as the EPI sequence (TR = 620 ms, TE = 12 ms, flip angle = 90 -, FOV = 240 mm, matrix = 224  256, Rect. FOV = 7/8, Effective Thickness = 1.25 mm), and a high resolution T1-weighted 3 D Magnetization Prepared Rapid Gradient Echo (MPRAGE) structural image was performed (TR = 11.4 ms, TE = 4.4 ms, flip angle = 8 -, FOV = 270 mm, matrix = 224  256, Rect. FOV = 7/8, Effective Thickness = 1.25 mm).

703

block design with the regressors entered as square waves (on = 1, off = 0), using a fixed effects general linear model as implemented in FSL. Each model was composed of the regressor indicating the task of interest, one for the instructions, the time derivatives of the two previous regressors, and regressors for head motion during the run. The task regressors were convolved with a canonical hemodynamic response function. The parameter estimates from each run were normalized to the Montreal Neurological Institute/International Consortium for Brain Mapping 152 standard (MNI/ICBM), as contained within the FSL software package. The normalization was calculated using the structural images of each subject. The first step was to remove the non-brain tissue in the structural images with BET [72] and then we manually edited the resulting images for any remaining non-brain tissue. The EPI images (after time slice correction and motion correction) were co-registered to the 28-slice T1 weighted image (7-parameter rigid body), then the 28-slice T1 weighted image was registered to the MPRAGE image (7-parameter rigid body), and then the MPRAGE image was registered to the MNI/ICBM template (12 parameter affine transformation). The parameter estimates from each run were transformed into the MNI/ICBM stereotaxic space for group analysis using the transformations described above. The parameter estimates and variance for each run were combined within the mixed effects model to calculate the group statistical maps as well as the contrasts between the two difficulty levels [6], see Fig. 2 for the design matrix. The voxel-based statistical analyses were based on a voxel wise threshold of Z = 2.33 ( P < 0.01) and each cluster was corrected for multiple comparisons using Gaussian random theory at the P < 0.05 level as implemented in FSL [23,26,81]. The group activation maps for the baseline and difficult level were calculated using a one sample t test. The contrast between the difficulty levels was calculated using a

2.4. Data analysis The data were analyzed off line in a computer with an Intel Pentium III CPU (San Jose, California, USA) with Linux (Red Hat version 7.0, Red Hat Inc, Rayleigh, North Carolina, USA) as the operating system and we utilized AFNI [18] (available at afni.nimh.nih.gov/afni/) and FSL (FMRIB Software Library-available at www.fmrib.ox.ac.uk/ fsl) for analysis. The initial step was to delete the first 4 volumes of each scan to remove the initial T1 magnetic transients in the data. The data were corrected for the time acquisition differences between slices using Fourier interpolation. Then the data were corrected for head motion (6parameter rigid body) where the reference volume was in the center of the run. The data were smoothed (Gaussian filter full width at half maximum = 8  8  8 mm) and high pass filtered with a cutoff at (1/100) Hz. The parameters estimates of each run were analyzed in a

Fig. 2. Design matrix (abbreviated) for the mixed effects voxel-based analysis showing in the first column the covariate for the task difficulty, and from the second column onwards the covariate for the average (mean) of the easy and difficult runs of each subject.

704

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

Table 1 Mean accuracies and response times for the matching tasks for each degradation category and combined into the two categories Task

Accuracy (%)

Response time (s)

Face Face + 10% Face + 20% Face + 40% Face + 60% Baseline level Difficult level

93.4 93.5 82.5 82.7 82.6 93.5 82.6

1.15 1.12 1.22 1.35 1.31 1.13 1.29

(9.3) (5.0) (5.1)* (8.9)* (12.4)* (7.6) (8.5)**

(0.32) (0.34) (0.41) (0.36) (0.27) (0.32) (0.36)

Data are mean and standard deviation in parentheses. * Statistically significantly different from the Face condition (and Face + 10%) at the P < 0.05 level. ** Statistically significantly different from the Baseline condition at the P < 0.0001 level.

paired t test. The peaks of activation were located with reference to the Talairach and Tournoux template [74]. To convert the MNI/ICBM coordinates to the Talairach and Tournoux coordinates, we utilized a non-linear transformation developed by M. Brett between both stereotaxic spaces (see online at http://www.mrc-cbu.cam.ac.uk/Imaging/ mnispace.html). The ROIs were in four regions in the right hemisphere: (a) the FFA, (b) the DLPFC, (c) the ACC and (d) the anterior fusiform gyrus. We defined the FFA ROI based on the average coordinates from other studies [15,20,28,36– 38,43,45,77,80]. The location in the right hemisphere was x, 40 mm; y, 54 mm; z, 12 mm. The ROI was a sphere with a radius of 5 mm. In addition, the right hemisphere DLPFC ROI was defined as consisting of all activated voxels within the middle frontal gyrus [20]. The ACC was defined as voxels in Brodmann Area (BA) 24 and 32 [74]. The anterior fusiform gyrus ROI was a sphere centered at x, 40 mm; y, 30 mm; z, 12 mm with a radius of 5 mm. The ROIs were drawn in the ICBM/MNI stereotaxic space and were transformed to the native space of each subject using the inverse of the transformations between the original FMRI data and the MNI/ICBM stereotaxic space. The statistical comparisons of the ROI data between the baseline and difficult condition were based on the delta magnitude ((baseline-control) (difficult-control)). Single sided paired t test corrected for multiple comparisons at the P < 0.05 level were used for comparing the delta magnitude differences between baseline and difficult condition.

and the 20%, 40% and 60% conditions. Based on the behavioral data, the data between the different degradation conditions were divided into two levels—the baseline level composed of the non-degraded condition and 10% degradation condition and the difficult level composed of the 20%, 40% and 60% degradation conditions. This was done because the accuracy level indicated that the various degradation levels did not linearly increase difficulty level, as we expected, but instead fell into two levels. Thus we divided the tasks into a baseline level and the difficult level. The accuracy was 93.5% (7.6%) and 82.6% (8.5%) for the baseline and difficult levels, respectively. The accuracy was significantly different at the P < 0.0001 level. The difference in response time between task difficulty levels (baseline versus difficult) was not statistically significant. When comparing the reaction time data for the 20% condition compared to 40% (or 60%) condition found no statistically significant difference (at the P < 0.05 level). Similarly, there was no statistically significant difference between the 20% condition and the baseline level (or either of the two conditions that form the baseline level). 3.2. Region of interest results There were significant differences in the magnitude of activation between the baseline and difficult level in the FFA (single sided paired t test, P < 0.0001, Bonferroni correction for multiple comparisons) (see Table 2). In addition, there was a statistically significant linear correlation between FFA and task performance (correct response rate) (cc = 0.367, P < 0.005) across all task difficulty levels. When re-analyzing the data with the response time as a covariate in an ANCOVA model, the differences in activation between the baseline and difficult condition remained statistically significant. The response time was not statistically significant with the activation level ( P = 0.07). We found a statistically significant increase in activation between the baseline and difficult level in the DLPFC (single sided paired t test, P < 0.05, Bonferroni correction for multiple comparisons). The correct response rate was not significantly correlated to the Table 2 Mean activation of the regions of interest located in the face fusiform gyrus, the dorsolateral prefrontal cortex, the anterior cingulate cortex and anterior fusiform gyrus of the right hemisphere

3. Results

Region

Baseline

Difficult

3.1. Behavioral results

Face fusiform area Dorso-lateral prefrontal Anterior cingulate cortex Anterior fusiform gyrus

0.40 0.55 0.13 0.28

0.61 0.75 0.26 0.26

The mean response accuracy and response time for each stimulus category were presented in Table 1. There were no differences in response time between the non-degraded and 10% condition but there was a statistically significant difference between the non-degraded (or 10% condition)

(0.19) (0.31) (0.19) (0.18)

(0.19)** (0.26)* (0.24) (0.25)

Mean values (standard deviation) of % increase of the BOLD signal. All statistical comparisons are corrected for multiple comparisons (Bonferroni correction). * Significant at the P < 0.01 level. ** Significant at the P < 0.0001 level.

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

705

Fig. 3. Statistical parametric maps of (A) statistically significant activation for the baseline level, (B) statistically significant increased activation due to task difficulty level, and (C) statistically significant decreased activation due to task difficulty. All maps are mixed effects with a voxel wise threshold of Z = 2.3 and a spatial correction at P < 0.05.

DLPFC. The data were reanalyzed using an ANCOVA model to include the response time as a covariate. The differences in activation between the baseline and difficult condition remained statistically significant, with the response time not modeling a statistically significant variance of the activation ( P = 0.40). The activation in the ACC ROI was not linearly correlated to task performance. In addition, the difference in activation between the baseline and difficult level was not statistically significant ( P < 0.05, Bonferroni correction). The control ROI in the anterior fusiform gyrus did not have increased activation between the baseline and difficult level (paired t test, P = 0.34), and the activation was not linearly correlated to performance ( P = 0.38). 3.3. Voxel-based imaging results The activation peaks during the baseline perceptual matching task compared to the control task (see Fig. 3A and Table 3) were located in the left fusiform gyrus, left middle occipital gyrus and bilaterally in the inferior occipital gyrus. In the right frontal lobe, the inferior and middle frontal gyri were activated. The activation peaks in the difficult perceptual matching task compared to the control condition were located in the left fusiform gyrus and left inferior frontal gyrus (Table 4).

Within the right hemisphere, the highest peaks of activation were located in the fusiform gyrus, inferior occipital gyrus, amygdala, inferior parietal lobule, inferior frontal gyrus, middle frontal gyrus and medial frontal gyrus. The contrast comparing the increase in activation during the difficult level to the baseline level led to relative increases in the right frontal lobe. The activation peaks were located within the inferior (BA 44, 45), middle frontal (BA 10) and superior frontal (BA 10) gyri and (Table 5 and Fig. Table 3 Location of highest activation peaks for the baseline level Region Left hemisphere Occipital lobe Inferior occipital gyrus Middle occipital gyrus Fusiform gyrus Right hemisphere Occipital Inferior occipital gyrus Frontal Inferior frontal gyrus Middle frontal gyrus

Brodmann area

x

y

z

Z value

18 19 37

40 44 42

84 66 63

10 9 14

8.40 8.36 8.15

18

24

101

5

7.35

45 44 6

52 44 46

40 15 7

5 23 53

7.16 8.41 5.38

706

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

Table 4 Location of highest activation peaks for the difficult level Region Left hemisphere Occipital lobe Fusiform gyrus Frontal lobe Inferior frontal gyrus

Right hemisphere Occipital lobe Fusiform gyrus Inferior occipital gyrus Temporal lobe Amygdala Parietal lobe Inferior parietal lobule Frontal lobe Inferior frontal gyrus

Middle frontal gyrus Medial frontal gyrus

Brodmann area

x

y

z

Z value

18 36

42 42

64 46

11 20

11.01 9.02

44 47

42 33

9 23

25 8

7.47 7.73

37 18

44 30

59 90

16 2

11.25 7.86

22

3

17

6.17

40

30

52

43

7.08

47 44 45 10 46 8

36 44 44 42 46 2

22 11 30 45 46 29

11 25 15 13 20 37

8.64 9.10 8.38 6.70 6.70 8.15

3B). In addition, when the response time was added as a covariate in the mixed effects model, the increase in activation in the frontal lobes in the difficult level compared to the baseline level did not remain statistically significant. There were no areas that were correlated with the response time. The increase in task difficulty led to decreased activation in the visual cortex in the occipital and temporal lobes (Table 6 and Fig. 3C).

4. Discussion The main findings from our study were the FFA was modulated by task difficulty and activation in other regions of the visual extrastriate was suppressed by task difficulty. We also found step changes in behavioral performance. In the voxel-based analysis, we also found increased activation in right inferior and middle frontal

gyri. In addition, task performance was correlated to activation in the FFA but not the DLPFC. There was no increase in activation in the DLPFC between task difficulty levels when response time was included as a covariate in the voxel-based analysis, but in the ROI analysis the differences in activation between difficulty levels remained statistically significant after including the response time covariate. 4.1. Behavioral results The behavioral response to increased image degradation was not linear as a previous study found [30]. There was a step-wise decrease in performance between the 10% and 20% degradation. The possible reason for the behavioral differences may be due to (a) the current task had 2 faces instead of three faces in each trial; (b) the faces were shown from the same perspective in all cases. These differences simplified the cognitive task compared to when three faces were presented [30]. Based on the behavioral performance, the participants did not perceive a difference between the non-degraded images and the 10% degraded images. Similarly, the differences among the 20%, 40% and 60% degraded images did not produce an increase in task difficulty among the three conditions. Thus, increasing image degradation does not necessarily lead to linear increases in task difficulty but may be non-linear. 4.2. Modulation of the face fusiform area In the ROI analysis we found that the FFA was modulated by task difficulty with increased activation as task difficulty increased. Additional supporting evidence that FFA was specifically modulated by task difficulty was the finding that performance was linearly correlated to FFA activation. Our findings are supported by a study of single cell recordings in non-human primates, which

Table 6 Local maxima of the decrease in the BOLD signal during the difficult level compared to the baseline level Region

Table 5 Local maxima of the increase in the BOLD signal during the difficult level compared to the baseline level Region Right hemisphere Inferior frontal gyrus

Brodmann area

x

45

44 50 44 44 40 38 30

44 Middle frontal gyrus

10

Superior frontal gyrus

10

y

34 26 13 8 52 49 53

z

15 4 29 18 9 3 3

Z value

3.75 3.11 3.38 3.21 2.85 2.80 2.97

Left hemisphere Cuneus Hippocampus Posterior cingulate Parahippocampal gyrus Right hemisphere Posterior cingulate Hippocampus

Brodmann area 18 17 23/30 31 30

29 23

x

y

z

Z value

8 4 18 16 2 22

73 83 35 54 55 39

17 4 2 14 28 5

3.69 3.81 2.77 3.97 3.43 2.46

2 14 28 36

46 52 35 28

10 12 4 10

3.24 3.43 2.88 3.31

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

found that neuronal activation in area V4 of the visual extrastriate was enhanced when task difficulty was increased in a visual discrimination task [73]. The present results extend previous findings that FFA activation was enhanced with increased working memory load using faces as stimuli [19]. Attention has also enhanced activation in specialized processing areas like the FFA [60], the parahippocampal place area [21,22] and in areas of the extrastriate visual cortex when the stimuli were presented within the receptive field of the specific area [46,47]. These findings, together with our results, suggest that visual extrastriate areas such as FFA were enhanced when task difficulty increased due to image degradation. The ROI analysis of the FFA showed increased sensitivity compared to the voxelwise method to detect changes in the BOLD signal between the baseline and difficult conditions. The increased sensitivity is at the cost of loss of specificity, that is, the location of the activation changes are not as specific as with the voxel-wise method. The ROI analysis is the average activation in a sphere (radius 5 mm) while the basic unit in the voxel-based analysis (in the standard stereotaxic space) is a cube with a length of 2 mm. In addition, previous studies that have mapped the FFA detected an activation region in some subjects composed of only a few voxels. The voxel-based analysis is not as sensitive compared to the ROI methodology to detect these very small regional changes. 4.3. Modulation of the dorsolateral prefrontal cortex and anterior cingulate cortex The increased activation in the DLPFC ROI may have indicated increased demand upon working memory as the two stimuli were compared, increased coordination among the regions activated in the task, and the increased visual search time of the degraded images. The activation of the frontal lobes has been shown to be involved in mediating attention, working memory, cognitive control in a wide range of studies (see for example [2,9,14,31, 32,41,44,49,55]). The activation in the baseline and difficult level were consistent with previous studies investigating face matching tasks [16,17,20,34,35,38] and tasks involving parametric task difficulty [2,30,33, 39,40]. In the ACC, we did not find any changes in activation due to task difficulty or linear correlation with task performance. The modulation found in the FFA and DLPFC may be due to non-specific effects such as arousal. Thus to confirm that the modulation was not due to non-specific effects, an ROI was located in the anterior fusiform gyrus anterior. The lack of significant increases in activation in the anterior fusiform gyrus ROI between the baseline and difficult level and lack of significant correlation between activation level and performance support the interpretation that task difficulty specifically modulated the FFA and DLPFC.

707

4.4. Voxel-based analysis The regions activated in the voxel-based analysis in both the baseline and difficult level included the visual areas in the posterior cortex, and the right inferior and middle frontal gyri. A proposed model of the human system for face perception suggests that the inferior occipital gyri are involved in early feature perception, while the fusiform area is involved in perception of invariant features of faces [38]. The regions of largest activation in the posterior cortex that we found in the present study were consistent with the proposed specialized face network. The processing from the face perception network provides the input for later stages of processing that occur, for example in the frontal lobes. The activation within the frontal lobes was localized to the right frontal cortex (inferior and middle frontal gyri) in the baseline level and bilaterally in the inferior and middle frontal gyri for the difficult level. In the voxel-based analysis we found that there was an increased in activation between the baseline and difficult conditions in the DLPFC but this difference disappeared when the response time was added as a covariate in the analysis. When adding the response time as a covariate in the re-analysis of the ROI data, the statistically significant differences in activation between the baseline and difficult level remained. One possibility for the difference may be that the large size of the DLPFC ROI may make is less sensitive to changes due to response time. The response time covariate may have correlated with relatively small regions in the middle frontal gyrus, thus in the ROI analysis, which is an average of the gyrus activation, may have made it less sensitive to the variance related to the response time covariate. In the voxel-based analysis, an additional area of activation in the anterior middle frontal gyrus was found. Activation within the anterior section of the middle frontal gyrus (BA 10) has been seen with difficult reasoning and problem solving tasks [7,29,58,63], in tasks requiring task switching and branching between different goals [50], task involving analogous relationships [79], and tasks involving relational complexity [51]. Activation within this region may reflect that degraded images place greater demands on reasoning and integration of complex relations between the stimuli. Within the constraints of the current study, it may be that the degraded stimuli was integrated with knowledge about the structure of faces into a useful representation for performance of the task. Tasks involving relationships, reasoning, integration of information generally recruited anterior areas of the middle frontal gyrus (BA 10). The decrease in activation (as task difficulty increased) within the visual cortex was consistent with the decreases found by in visual match-to-sample task [30] and in tone matching task [39]. Grady et al. [30] suggested that the decrease in activation within the visual cortex is due to increased reliance upon the frontal lobes for performance of

708

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710

the task. Our results indicated that there was increased activation in the FFA for processing of visual stimuli and in parallel there was suppression of activation in other areas of the visual cortex. In addition, the increase in activation in the DLPFC due to task difficulty did not remain after taking into account response time in the voxel-based analysis but the difference did remain in the ROI analysis. The differences between both analysis methods may be due to differences in sensitivity as explained previously. The differences in performance between the baseline and difficult level could have had an effect on the activation differences that we found due to the different number of movements (responses). In the difficult condition there were fewer movements (pressing the button). This could lead to differences in activation in the motor cortex with higher activation during the baseline condition. Given that we did not find this difference, it is unlikely that it had a significant effect upon the activation results. 4.5. Relationship between BOLD signal and neuronal activation The causal relationship between the BOLD signal and local brain activity has been shown to be linear when utilizing the local field potential as the measure of neuronal activation [54]. The local field potential is a measure of neuronal input into a region and of the local processing of neuronal information within a region. Thus, the increased BOLD signal found between the tasks difficulties most likely reflected the changes in neuronal activation locally as well as the input signal into the region. In addition, it has been shown that the spatial pattern of neuronal activation matched the spatial pattern of the BOLD signal [48,83]. The BOLD signal also reflects changes in cerebral blood flow, cerebral blood volume, and oxygen consumption. The interaction between these variables and the BOLD signal involve various factors such as cell types, circuitry driven during activation and the relationship that couple energy demand to its supply to the brain. The temporal dynamics of the BOLD signal and its relationship to cerebral blood flow, oxygen consumption and blood volume changes are nonlinear [1,11,12,57,59,76,84]. In general, the relationship between stimulus energy and BOLD response is non-linear, with the non-linearities present in the step from stimulus energy to neuronal activation [42,64 – 66,75] and from neuronal activation to cerebral blood flow [3,24,25,52, 53,68,70] and BOLD response [78]. Within a certain range of this function, the response between neuronal activation and BOLD response is linear [10,13,56,61,62,69]. 4.6. Other modulatory processes Drugs can also modulate neuronal activation as shown in a group of young healthy subjects performing a working memory task [27]. When the subjects received an infusion of physostigmine there was an increase in cerebral blood

flow in the visual areas and decrease in cerebral blood flow in the frontal areas compared to a saline infusion. Furey et al. [27] found that physostigmine may have improved performance by enhancing visual processing of the stimuli. The present study can be extended in an event-related design where task difficulty is parametrically varied from trial to trial. Another issue that may merit further investigation is if different methods of stimuli degradation lead to similar response modulation, for example degrading the images by removing the high or low frequency components of the faces. In addition, the effects of task difficulty can be investigated in patient populations.

5. Conclusions The increase in task difficulty led to enhanced selectivity of the FFA which is involved in processing the face stimuli. In addition, there was a decrease in activation within other visual extrastriate areas. Task difficulty modulated regions in the visual extrastriate cortex differentially.

Acknowledgments The authors thank P. Lopez-Bayo, S. Pechler and M. Karman for their assistance in acquiring and analyzing the data. This study was supported by a grant from the Volkswagen Stiftung (Hannover, Germany) to A.B., S.J.T. and H.H.

References [1] A. Aubert, R. Costalat, H. Duffau, H. Benali, Modeling of pathophysiological coupling between brain electrical activation, energy metabolism and hemodynamics: insights for the interpretation of intracerebral tumor imaging, Acta Biotheor. 50 (2002) 281 – 295. [2] D.M. Barch, T.S. Braver, L.E. Nystrom, S.D. Forman, D.C. Noll, J.D. Cohen, Dissociating working memory from task difficulty in human prefrontal cortex, Neuropsychologia 35 (1997) 1373 – 1380. [3] S.H. Bartley, Temporal features of input as crucial factors in vision, in: W.D. Neff (Ed.), Contrib. Sens. Physiol., vol. 3, Academic Press, New York, 1968, pp. 81 – 124. [4] G.C. Baylis, E.T. Rolls, C.M. Leonard, Selectivity between faces in the responses of a population of neurons in the cortex in the superior temporal sulcus of the monkey, Brain Res. 342 (1985) 91 – 102. [5] G.C. Baylis, E.T. Rolls, C.M. Leonard, Functional subdivisions of the temporal lobe neocortex, J. Neurosci. 7 (1987) 330 – 342. [6] C.F. Beckmann, M. Jenkinson, S.M. Smith, General multilevel linear modeling for group analysis in FMRI, NeuroImage 20 (2003) 1052 – 1063. [7] K.F. Berman, J.L. Ostrem, C. Randolph, J. Gold, T.E. Goldberg, R. Coppola, R.E. Carson, P. Herscovitch, D.R. Weinberger, Physiological activation of a cortical network during performance of the Wisconsin Card Sorting Test: a positron emission tomography study, Neuropsychologia 33 (1995) 1027 – 1046. [8] V. Blanz, T. Vetter, A morphable model for the synthesis of 3 D faces, 26th INternational Conference on Computer Graphics and Interactive Techniques, ACM, Los Angeles, California, USA, 1999.

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710 [9] T.S. Braver, J.D. Cohen, L.E. Nystrom, J. Jonides, E.E. Smith, D.C. Noll, A parametric study of prefrontal cortex involvement in human working memory, NeuroImage 5 (1997) 49 – 62. [10] G. Brinker, C. Bock, E. Busch, H. Krep, K.A. Hossmann, M. HoehnBerlage, Simultaneous recording of evoked potentials and T2*weighted MR images during somatosensory stimulation of rat, Magn. Reson. Med. 41 (1999) 469 – 473. [11] R.B. Buxton, E.C. Wong, L.R. Frank, Dynamics of blood flow and oxygenation changes during brain activation: the balloon model, Magn. Reson. Med. 39 (1998) 855 – 864. [12] R.B. Buxton, K. Uludag, D.J. Dubowitz, T.T. Liu, Modeling the hemodynamic response to brain activation, NeuroImage 23 (2004) S220 – S233. [13] K. Caesar, N. Akgoren, C. Mathiesen, M. Lauritzen, Modification of activity-dependent increases in cerebellar blood flow by extracellular potassium in anaesthetized rats, J. Physiol. 520 (1999) 281 – 292. [14] J.H. Callicott, V.S. Mattay, A. Bertolino, K. Finn, R. Coppola, J.A. Frank, T.E. Goldberg, D.R. Weinberger, Physiological characteristics of capacity constraints in working memory as revealed by functional MRI, Cereb. Cortex 9 (1999) 20 – 26. [15] V.P. Clark, K. Keil, J.M. Maisog, S. Courtney, L.G. Ungerleider, J.V. Haxby, Functional magnetic resonance imaging of human visual cortex during face matching: a comparison with positron emission tomography, NeuroImage 4 (1996) 1 – 15. [16] M. Corbetta, F.M. Miezin, G.L. Shulman, S.E. Petersen, Selective attention modulates extrastriate visual regions in humans during visual feature discrimination and recognition, Ciba Found. Symp. 163 (1991) 165 – 175 (discussion 175 – 180). [17] M. Corbetta, F.M. Miezin, G.L. Shulman, S.E. Petersen, A PET study of visuospatial attention, J. Neurosci. 13 (1993) 1202 – 1226. [18] R.W. Cox, AFNI: software for analysis and visualization of functional magnetic resonance neuroimages, Comput. Biomed. Res. 29 (1996) 162 – 173. [19] T.J. Druzgal, M. D’Esposito, Activity in fusiform face area modulated as a function of working memory load, Brain Res. Cogn. Brain Res. 10 (2001) 355 – 364. [20] T.J. Druzgal, M. D’Esposito, A neural network reflecting decisions about human faces, Neuron 32 (2001) 947 – 955. [21] R. Epstein, N. Kanwisher, A cortical representation of the local visual environment, Nature 392 (1998) 598 – 601. [22] R. Epstein, A. Harris, D. Stanley, N. Kanwisher, The parahippocampal place area: recognition, navigation, or encoding? Neuron 23 (1999) 115 – 125. [23] S.D. Forman, J.D. Cohen, M. Fitzgerald, W.F. Eddy, M.A. Mintun, D.C. Noll, Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): use of a cluster-size threshold, Magn. Reson. Med. 33 (1995) 636 – 647. [24] P.T. Fox, M.E. Raichle, Stimulus rate dependence of regional cerebral blood flow in human striate cortex demonstrated by positron emission tomography, J. Neurophysiol. 51 (1984) 1109 – 1120. [25] P.T. Fox, M.E. Raichle, Stimulus rate determines regional blood flow in striate cortex, Ann. Neurol. 17 (1985) 303 – 305. [26] K.J. Friston, K.J. Worlsey, R.S.J. Frakowiak, J.C. Mazziotta, A.C. Evans, Assessing the significance of focal activations using their spatial extent, Hum. Brain Mapp. 1 (1994) 214 – 220. [27] M.L. Furey, P. Pietrini, J.V. Haxby, G.E. Alexander, H.C. Lee, J. VanMeter, C.L. Grady, U. Shetty, S.I. Rapoport, M.B. Schapiro, U. Freo, Cholinergic stimulation alters performance and task-specific regional cerebral blood flow during working memory, Proc. Natl. Acad. Sci. U. S. A. 94 (1997) 6512 – 6516. [28] I. Gauthier, M.J. Tarr, J. Moylan, P. Skudlarski, J.C. Gore, A.W. Anderson, The fusiform ‘‘face area’’ is part of a network that processes faces at the individual level, J. Cogn. Neurosci. 12 (2000) 495 – 504. [29] V. Goel, B. Gold, S. Kapur, S. Houle, The seats of reason? An imaging study of deductive and inductive reasoning, NeuroReport 8 (1997) 1305 – 1310.

709

[30] C.L. Grady, B. Horwitz, P. Pietrini, M.J. Mentis, L.G. Ungerleider, S.I. Rapoport, J.V. Haxby, Effect of task difficulty on cerebral blood flow during perceptual matching of faces, Hum. Brain Mapp. (1996) 227 – 239. [31] P.M. Grasby, C.D. Frith, K. Friston, R.S. Frackowiak, R.J. Dolan, Activation of the human hippocampal formation during auditory – verbal long-term memory function, Neurosci. Lett. 163 (1993) 185 – 188. [32] P.M. Grasby, C.D. Frith, K.J. Friston, C. Bench, R.S. Frackowiak, R.J. Dolan, Functional mapping of brain areas implicated in auditory— Verbal memory function, Brain 116 (1993) 1 – 20. [33] R.C. Gur, R.E. Gur, B.E. Skolnick, S.M. Resnick, F.L. Silver, J. Chawluk, L. Muenz, W.D. Obrist, M. Reivich, Effects of task difficulty on regional cerebral blood flow: relationships with anxiety and performance, Psychophysiology 25 (1988) 392 – 399. [34] J.V. Haxby, C.L. Grady, B. Horwitz, L.G. Ungerleider, M. Mishkin, R.E. Carson, P. Herscovitch, M.B. Schapiro, S.I. Rapoport, Dissociation of object and spatial visual processing pathways in human extrastriate cortex, Proc. Natl. Acad. Sci. U. S. A. 88 (1991) 1621 – 1625. [35] J.V. Haxby, B. Horwitz, L.G. Ungerleider, J.M. Maisog, P. Pietrini, C.L. Grady, The functional organization of human extrastriate cortex: a PET-rCBF study of selective attention to faces and locations, J. Neurosci. 14 (1994) 6336 – 6353. [36] J.V. Haxby, L.G. Ungerleider, B. Horwitz, J.M. Maisog, S.I. Rapoport, C.L. Grady, Face encoding and recognition in the human brain, Proc. Natl. Acad. Sci. U. S. A. 93 (1996) 922 – 927. [37] J.V. Haxby, L.G. Ungerleider, V.P. Clark, J.L. Schouten, E.A. Hoffman, A. Martin, The effect of face inversion on activity in human neural systems for face and object perception, Neuron 22 (1999) 189 – 199. [38] J.V. Haxby, E.A. Hoffman, M.I. Gobbini, The distributed human neural system for face perception, Trends Cogn. Sci. 4 (2000) 223 – 233. [39] H.H. Holcomb, B. Gordon, H.L. Loats, E. Gastineau, Z. Zhao, D. Medoff, R.F. Dannals, R. Woods, C.A. Tamminga, Brain metabolism patterns are sensitive to attentional effort associated with a tone recognition task, Biol. Psychiatry 39 (1996) 1013 – 1022. [40] H.H. Holcomb, D.R. Medoff, P.J. Caudill, Z. Zhao, A.C. Lahti, R.F. Dannals, C.A. Tamminga, Cerebral blood flow relationships associated with a difficult tone recognition task in trained normal volunteers, Cereb. Cortex 8 (1998) 534 – 542. [41] G.D. Honey, E.T. Bullmore, T. Sharma, Prolonged reaction time to a verbal working memory task predicts increased power of posterior parietal cortical activation, NeuroImage 12 (2000) 495 – 503. [42] B. Horwitz, Simulating functional interactions in the brain: a model for examining correlations between regional cerebral metabolic rates, Int. J. Biomed. Comput. 26 (1990) 149 – 170. [43] A. Ishai, L.G. Ungerleider, A. Martin, J.V. Haxby, The representation of objects in the human occipital and temporal cortex, J. Cogn. Neurosci. 12 (2000) 35 – 51. [44] J.M. Jansma, N.F. Ramsey, R. Coppola, R.S. Kahn, Specific versus nonspecific brain activity in a parametric N-back task, NeuroImage 12 (2000) 688 – 697. [45] N. Kanwisher, J. McDermott, M.M. Chun, The fusiform face area: a module in human extrastriate cortex specialized for face perception, J. Neurosci. 17 (1997) 4302 – 4311. [46] S. Kastner, P. De Weerd, R. Desimone, L.G. Ungerleider, Mechanisms of directed attention in the human extrastriate cortex as revealed by functional MRI, Science 282 (1998) 108 – 111. [47] S. Kastner, D.H. O’Connor, M.M. Fukui, H.M. Fehd, U. Herwig, M.A. Pinsk, Functional imaging of the human lateral geniculate nucleus and pulvinar, J. Neurophysiol. 17 (2003) 17. [48] D.S. Kim, I. Ronen, C. Olman, S.G. Kim, K. Ugurbil, L.J. Toth, Spatial relationship between neuronal activity and BOLD functional MRI, NeuroImage 21 (2004) 876 – 885. [49] T. Klingberg, B.T. O’Sullivan, P.E. Roland, Bilateral activation of

710

[50]

[51]

[52]

[53]

[54]

[55]

[56]

[57]

[58] [59]

[60]

[61]

[62]

[63]

[64] [65]

A.L.W. Bokde et al. / Cognitive Brain Research 25 (2005) 701 – 710 fronto-parietal networks by incrementing demand in a working memory task, Cereb. Cortex 7 (1997) 465 – 471. E. Koechlin, G. Basso, P. Pietrini, S. Panzer, J. Grafman, The role of the anterior prefrontal cortex in human cognition, Nature 399 (1999) 148 – 151. J.K. Kroger, F.W. Sabb, C.L. Fales, S.Y. Bookheimer, M.S. Cohen, K.J. Holyoak, Recruitment of anterior dorsolateral prefrontal cortex in human reasoning: a parametric study of relational complexity, Cereb. Cortex 12 (2002) 477 – 485. E. Leniger-Follert, Microflow and evoked potentials in the somatomotor cortex of the cat brain during specific sensory activation, Acta Neurol. Scand., Suppl. 72 (60) (1979) 10 – 11. M. Livingston, D. Hubel, Segregation of form, color, movement, and depth: anatomy, physiology, and perception, Science 240 (1988) 740 – 749. N.K. Logothetis, J. Pauls, M. Augath, T. Trinath, A. Oeltermann, Neurophysiological investigation of the basis of the fMRI signal, Nature 412 (2001) 150 – 157. D.S. Manoach, G. Schlaug, B. Siewert, D.G. Darby, B.M. Bly, A. Benfield, R.R. Edelman, S. Warach, Prefrontal cortex fMRI signal changes are correlated with working memory load, NeuroReport 8 (1997) 545 – 549. C. Mathiesen, K. Caesar, N. Akgoren, M. Lauritzen, Modification of activity-dependent increases of cerebral blood flow by excitatory synaptic activity and spikes in rat cerebellar cortex, J. Physiol. 512 (1998) 555 – 566. K.L. Miller, W.M. Luh, T.T. Liu, A. Martinez, T. Obata, E.C. Wong, L.R. Frank, R.B. Buxton, Nonlinear temporal dynamics of the cerebral blood flow response, Hum. Brain Mapp. 13 (2001) 1 – 12. P. Nichelli, J. Grafman, P. Pietrini, D. Alway, J.C. Carton, R. Miletich, Brain activity in chess playing, Nature 369 (1994) 191. T. Obata, T.T. Liu, K.L. Miller, W.M. Luh, E.C. Wong, L.R. Frank, R.B. Buxton, Discrepancies between BOLD and flow dynamics in primary and supplementary motor areas: application of the balloon model to the interpretation of BOLD transients, NeuroImage 21 (2004) 144 – 153. K.M. O’Craven, P.E. Downing, N. Kanwisher, fMRI evidence for objects as the units of attentional selection, Nature 401 (1999) 584 – 587. S. Ogawa, T.M. Lee, A.R. Kay, D.W. Tank, Brain magnetic resonance imaging with contrast dependent on blood oxygenation, Proc. Natl. Acad. Sci. U. S. A. 87 (1990) 9868 – 9872. S. Ogawa, T.M. Lee, R. Stepnoski, W. Chen, X.H. Zhu, K. Ugurbil, An approach to probe some neural systems interaction by functional MRI at neural time scale down to milliseconds, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 11026 – 11031. D. Osherson, D. Perani, S. Cappa, T. Schnur, F. Grassi, F. Fazio, Distinct brain loci in deductive versus probabilistic reasoning, Neuropsychologia 36 (1998) 369 – 376. T. Pellmar, Electrophysiological correlates of peroxide damage in guinea pig hippocampus in vitro, Brain Res. 364 (1986) 377 – 381. T.C. Pellmar, Histamine decreases calcium-mediated potassium current in guinea pig hippocampal CA1 pyramidal cells, J. Neurophysiol. 55 (1986) 727 – 738.

[66] T.C. Pellmar, D.A. Schauer, G.H. Zeman, Time- and dose-dependent changes in neuronal activity produced by X radiation in brain slices, Radiat. Res. 122 (1990) 209 – 214. [67] D.I. Perrett, E.T. Rolls, W. Caan, Visual neurones responsive to faces in the monkey temporal cortex, Exp. Brain Res. 47 (1982) 329 – 342. [68] M.E. Phelps, J.C. Mazziotta, D.E. Kuhl, M. Nuwer, J. Packwood, J. Metter, J. Engel Jr., Tomographic mapping of human cerebral metabolism visual stimulation and deprivation, Neurology 31 (1981) 517 – 529. [69] G. Rees, K. Friston, C. Koch, A direct quantitative relationship between the functional properties of human and macaque V5, Nat. Neurosci. 3 (2000) 716 – 723. [70] D. Regan, Evoked Potentials in Psychology, Sensory Physiology, and Clinical Medicine, Chapman and Hall, London, 1972, 328 pp. [71] E.T. Rolls, G.C. Baylis, Size and contrast have only small effects on the responses to faces of neurons in the cortex of the superior temporal sulcus of the monkey, Exp. Brain Res. 65 (1986) 38 – 48. [72] S.M. Smith, Fast robust automated brain extraction, Hum. Brain Mapp. 17 (2002) 143 – 155. [73] H. Spitzer, R. Desimone, J. Moran, Increased attention enhances both behavioral and neuronal performance, Science 240 (1988) 338 – 340. [74] J. Talaraich, P. Tournoux, Co-Planar Stereotaxic Atlas of the Human Brain, Thieme Medical, New York, 1988. [75] J.M. Tolliver, T.C. Pellmar, Ionizing radiation alters neuronal excitability in hippocampal slices of the guinea pig, Radiat. Res. 112 (1987) 555 – 563. [76] K. Uludag, D.J. Dubowitz, E.J. Yoder, K. Restom, T.T. Liu, R.B. Buxton, Coupling of cerebral blood flow and oxygen consumption during physiological activation and deactivation measured with fMRI, NeuroImage 23 (2004) 148 – 155. [77] P. Vuilleumier, J.L. Armony, J. Driver, R.J. Dolan, Effects of attention and emotion on face processing in the human brain: an event-related fMRI study, Neuron 30 (2001) 829 – 841. [78] B.A. Wandell, Computational neuroimaging of human visual cortex, Annu. Rev. Neurosci. 22 (1999) 145 – 173. [79] C.M. Wharton, J. Grafman, S.S. Flitman, E.K. Hansen, J. Brauner, A. Marks, M. Honda, Toward neuroanatomical models of analogy: a positron emission tomography study of analogical mapping, Cognit. Psychol. 40 (2000) 173 – 197. [80] E. Wojciulik, N. Kanwisher, J. Driver, Covert visual attention modulates face-specific activity in the human fusiform gyrus: fMRI study, J. Neurophysiol. 79 (1998) 1574 – 1578. [81] K.J. Worsley, A.C. Evans, S. Marrett, P. Neelin, A three-dimensional statistical analysis for CBF activation studies in human brain, J. Cereb. Blood Flow Metab. 12 (1992) 900 – 918. [82] S. Yamane, S. Kaji, K. Kawano, What facial features activate face neurons in the inferotemporal cortex of the monkey? Exp. Brain Res. 73 (1988) 209 – 214. [83] X. Yang, F. Hyder, R.G. Shulman, Functional MRI BOLD signal coincides with electrical activity in the rat whisker barrels, Magn. Reson. Med. 38 (1997) 874 – 877. [84] Y. Zheng, J. Martindale, D. Johnston, M. Jones, J. Berwick, J. Mayhew, A model of the hemodynamic response and oxygen delivery to brain, NeuroImage 16 (2002) 617 – 637.