Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Journal of the International Neuropsychological Society (2002), 8, 607–622. Copyright © 2002 INS. Published by Cambridge University Press. Printed in the USA. DOI: 10.1017.S1355617702801394 Semantic monitoring of words with emotional connotation during f MRI: Contribution of anterior left frontal cortex BRUCE CROSSON,1,7 M. ALLISON CATO,1 JOSEPH R. SADEK,1 DIDEM GÖKÇAY,2 RUSSELL M. BAUER,1 IRA S. FISCHLER,3 LEEZA MARON,1 KAUNDINYA GOPINATH,4 EDWARD J. AUERBACH,5 SAMUEL R. BROWD,6 and RICHARD W. BRIGGS 5 McKnight Brain Institute of the University of Florida and Departments of 1 Clinical and Health Psychology, 2 Computer Information Sciences and Engineering, 3 Psychology, 4 Nuclear and Radiological Engineering, 5 Radiology, 6 Neuroscience University of Florida Health Science Center, Gainesville, Florida 7 Veterans Affairs RR&D Brain Rehabilitation Research Center, VA Medical Center, Gainesville, Florida (Received January 26, 2001; Revised August 27, 2001; Accepted September 5, 2001) Abstract Previous studies showed that cortex in the anterior portions of the left frontal and temporal lobes participates in generating words with emotional connotations and processing pictures with emotional content. If these cortices process the semantic attribute of emotional connotation, they should be active whenever processing emotional connotation, without respect to modality of input or mode of output. Thus, we hypothesized that they would activate during monitoring of words with emotional connotations. Sixteen normal subjects performed semantic monitoring of words with emotional connotations, animal names, and implement names during f MRI. Cortex in the anterior left frontal lobe demonstrated significant activity for monitoring words with emotional connotations compared to monitoring tone sequences, animal names, or implement names. Together, the current and previous results implicate cortex in the anterior left frontal lobe in semantic processing of emotional connotation, consistent with connections of this cortex to paralimbic association areas. Current findings also indicate that neural substrates for processing emotional connotation are independent of substrates for processing the categories of living and nonliving things. (JINS, 2002, 8, 607– 622) Keywords: Frontal lobe, Temporal lobe, Limbic system, Functional magnetic resonance imaging, Emotion, Semantics tinguishing amongst living and nonliving things, respectively (e.g., Hart & Gordon, 1992; Warrington & McCarthy, 1983). Cortex in the ventral temporal lobe, referred to as the “ventral visual stream,” is critical for analyzing visual form (Ungerleider & Mishkin, 1982) and frequently is damaged in patients having category-specific deficits for living things (e.g., Ferreira et al., 1997; Warrington & Shallice, 1984). However, functional neuroimaging experiments have produced conflicting results regarding the role of the ventral visual stream in processing visual form as a semantic attribute. Thompson-Schill et al. (1999) asserted that processing verbal information about living items engaged cortex in the fusiform gyrus (i.e., in the ventral visual stream) because of the importance of visual form in defining living things, but verbal processing of nonliving items engaged cortex of the fusiform gyrus only when subjects attended to INTRODUCTION Attributes are features that distinguish members of semantic categories from each other. Inability to process particular semantic attributes after brain injury may contribute to category-specific semantic impairments. The most common category-specific deficits involve living versus nonliving things, or sometimes more specifically animals versus implements (e.g., Caramazza & Shelton, 1998; Damasio et al., 1996; Hillis & Caramazza, 1991; Sacchett & Humphreys, 1992; Tranel et al., 1997). The semantic attributes of visual form and function are thought important for dis- Reprint requests to: Bruce Crosson, Ph.D., Department of Clinical and Health Psychology, University of Florida Health Science Center, Box 100165, Gainesville, FL 32610-0165. E-mail: [email protected] 607 608 visual form. On the other hand, Chao et al. (1999) demonstrated that the medial fusiform gyrus is more sensitive to nonliving things, and the lateral fusiform gyrus to living things, even when items were presented as written names and no attempt was explicitly made to analyze visual form. One problem with the latter study was that differences in location of activity for nonliving and living things (Chao et al., 1999) could occur either because their visual attributes required separate perceptual mechanisms, or because a visual semantic system in the fusiform gyrus is topographically divided by category. Indeed, findings of previous studies have supported the existence of a separate visual and verbal semantic systems (e.g., Hart & Gordon, 1992; McCarthy & Warrington, 1988). Because cortex in the ventral visual stream could play a fundamental role in object identification (i.e., visual semantics) in addition to processing visual attributes, attributes other than visual form must be studied to ascertain the existence of neural substrates for semantic attributes apart from the distinction of semantic categories. Emotional connotation is an ideal semantic attribute to investigate neural substrates. First, emotional connotation varies between objects within traditional semantic categories, such as living and nonliving things. For example, within the category of nonliving things, a cake has a positive emotional connotation, a coffin has a negative emotional connotation, and hammer has a neutral emotional connotation. This feature of emotional connotation allows investigators to rule out common semantic categories as a confounding factor in the study of emotional connotation. Second, our knowledge of emotional processing systems in the brain allows us to identify cortical regions that have access to emotional information, which should be good candidates for processing the semantic attribute of emotional connotation. Thus, cortices with access to the limbic system are good candidates for processing emotional connotation. Indeed, cortices in the more anterior portions of the left frontal and temporal lobes, which have access to limbic structures and0or paralimbic association areas (Pandya & Yeterian, 1986), show increased activity when subjects generate words with emotional connotations (Crosson et al., 1999). Activity in the more anterior portions of the right or both temporal lobes (Canli et al., 1998; Lane et al., 1999) and in the more anterior portion of the left frontal lobe (Lane et al., 1997; Paradiso et al., 1999) appears when viewing pictures with emotional content. Not only are these cortices characterized by connections to limbic structures, they are outside areas usually implicated in the more central aspects of either visual or verbal semantic processing. Thus, activity in the more anterior portions of the left frontal and temporal lobes cannot be attributed to fundamental semantic processes in either the visual or verbal modalities. However, while these findings regarding generating words and viewing pictures with emotional connotations suggest that cortices in the anterior left frontal and temporal lobes are involved in processing the semantic attribute of emotional connotation, one other criterion must be met in order B. Crosson et al. to confirm this hypothesis: The comprehension of words with emotional connotations must demonstrate involvement of the same cortices. Considerable evidence indicates that spoken words, heard words, and seen objects may have separate lexical or visual form representations (e.g., Ellis & Young, 1988). Each of these lexical and visual forms access stored semantic information to understand or to convey meaning. If the contribution of a cortical structure is based on representation of a particular kind of lexical or visual form then it should be uniquely active when that kind of lexical or visual form is processed. On the other hand, if the contribution of a cortical structure is based on the representation of semantic information, it will be engaged whenever the specific kind of semantic representations that it supports are processed. Thus, cortices that process the semantic attribute of emotional connotation should be active whenever emotional connotation is processed, without respect to the source of lexical or visual input, or without respect to mode of output. It follows that if cortices in the anterior left frontal and temporal lobes are important for processing emotional connotation as a semantic attribute, they should demonstrate increased activity when recognizing spoken words with emotional connotations, as well as when generating words with emotional connotations or viewing pictures with emotional content. To test this hypothesis, neurologically normal subjects performed three semantic monitoring tasks during f MRI of left-hemisphere activity: monitoring of nouns with emotional connotations, monitoring of implement names, and monitoring of animal names. The four-spiral pulse sequence which we used to acquire our data allowed us to collect only nine functional imaging slices, which precluded a whole-brain acquisition. We chose to cover the entire left hemisphere with these nine slices because of the dominance of the left hemisphere for processing words, with or without emotional connotations. For example, lesion data suggest that difficulty processing emotional connotation through linguistic channels is a function of linguistic competence of the left hemisphere rather than a function of the right hemisphere (Heilman et al., 1993). Further, the data of Ali and Cimino (1997) indicate that words tachistoscopically presented to the left hemisphere (right visual field) are more accurately perceived than words presented to the right hemisphere (left visual field) for positive, negative, or neutral emotional connotation. The magnitude of this effect is greater than within hemisphere differences in accuracy between word types. Each of the three semantic monitoring tasks in the present experiment was alternated with a tone monitoring task during separate imaging runs. It was hypothesized that monitoring of nouns with emotional connotations would demonstrate greater activity than the tone monitoring control task in the more anterior portions of the left frontal and temporal lobes, while monitoring of implement or animal names would not show such changes in comparison to the tone monitoring control task. One weakness in our previous study of generation of words with emotional connotations (Crosson et al., 1999) was that no Semantic monitoring of emotional connotation systematic attempt was made to exclude the distinction between living and nonliving categories as a potential confound. Thus, in the current study, words with emotional connotations equally represented living and nonliving things, and items for the two semantic monitoring comparison tasks were chosen from prototypical living items (animal names) and prototypical nonliving items (tool names). If the processing of emotional connotation in the more anterior portions of the left frontal and temporal lobes is orthogonal to the living–nonliving distinction, processing of emotional connotation should show greater activity at these locations than monitoring of either implement or animal names. Finally, since words in all three semantic monitoring tasks were selected for their ability to evoke visual imagery, we hypothesized all semantic monitoring tasks would show activity in the left fusiform gyrus compared to the tone monitoring task. MATERIALS AND METHODS Research Participants Sixteen strongly right-handed healthy participants (8 male, 8 female) took part in the study. Subjects’ ages ranged from 21 to 37 years (M 5 26.3 years, SD 5 5.0); education ranged from 12 to 24 years (M 5 17.7 years, SD 5 3.0). Edinburgh Handedness Inventory laterality quotients (Oldfield, 1971) ranged from 53 to 100 (M 5 82.8, SD 5 18.6), indicating strong right handedness. All participants gave written informed consent according to procedures established by the Health Center Institutional Review Board at the University of Florida. Experimental Tasks and Manipulation Check Participants alternated between a semantic monitoring task and a tone monitoring task during three separate f MRI imaging runs. In the semantic monitoring tasks, subjects monitored one specific type of word for two relevant semantic characteristics. During the control task, subjects monitored tone sequences for the presence of two or more highpitched tones. Each functional imaging run consisted of 8 cycles of alternation between semantic monitoring and tone monitoring. Half cycles lasted 28 s, and stimuli were presented at precise intervals using a computerized playback system. Stimuli were presented binaurally with volume individually adjusted for optimal hearing during scanning (Kenwood KR-A4070 amplifier, Realistic 31-2005 Ten band stereo frequency equalizer, JBL 16 W speaker, and acoustically insulated air conduction transducer with foam insert earphones). Stimuli in the semantic monitoring tasks were spoken English nouns. Nine spoken English nouns were presented in each 28-s half cycle, for a total of 72 nouns across all 8 cycles. Three types of words were presented in separate experimental runs: words with emotional connota- 609 tions, animal names, and implement names. No words were duplicated between or within the different experimental runs. To ensure the presence of implied emotional content, nouns with emotional connotations were selected from the Affective Norms for Emotional Words (ANEW; Bradley et al., 1988). ANEW words were rated for emotional valence and arousal by a large sample of normal participants similar to participants from the current study. Ratings for emotional valence were made on a 9-point scale with 1 representing the most emotionally negative valence, 9 representing the most emotionally positive valence, and 5 representing emotionally neutral. ANEW ratings for emotional arousal were made on a 9-point scale, with 1 representing minimally arousing, 9 representing maximally arousing, and 5 representing a moderate level of arousal. Using the valence and arousal ratings from ANEW, nouns with negative connotations were chosen from words with valence ratings less than 3.5 and with arousal ratings above 5. Nouns with positive connotations were selected from words with valence ratings higher than 6.5 and arousal ratings above 5. All nouns were selected for high imageability (above 3, range 1–5) because stimuli for the comparison tasks, animal and tool names, were highly imageable. The purpose of the present experiment was to study lefthemisphere activity during comprehension of a semantic attribute, emotional connotation, which was orthogonal to common semantic categories. The most common categoryspecific deficit after brain damage is living and nonliving things (e.g., Caramazza & Shelton, 1998; Damasio et al., 1996; Hillis & Caramazza, 1991; Sacchett & Humphreys, 1992; Tranel et al., 1997). Thus, to ensure that emotional connotation would not be confounded with the distinction between living and nonliving categories, words with emotional connotations consisted of equal numbers of living and nonliving items. Further, since our intention was to study emotional connotation as a general semantic attribute, equal numbers of words with positive emotional valence and negative emotional valence were selected. To further ensure that processing of emotional connotation was independent of the living-nonliving distinction, protypical living items (animals names) and nonliving items (implement names) were selected for semantic monitoring comparison tasks. Based on these considerations, equal numbers of words with emotional connotations were selected from the following possibilities: positive emotional connotation–living, positive emotional connotation–nonliving, negative emotional connotation–living, and negative emotional connotation– nonliving. There was no overlap between the list of words with emotional connotations and the animal or implement lists. Equal numbers of animal names were selected from the following possibilities: water dwelling–two legs or less; water dwelling–more than two legs; land dwelling–two legs or less; land dwelling–more than two legs. Implement names were equally selected from the following groups: used primarily outdoors–requires two hands; used primarily outdoors–requires one hand only; used primarily indoors– 610 requires two hands; used primarily indoors–requires one hand only. Because animal names generally are less commonly used than tool names in English, because all words with emotional connotation were selected from the ANEW list, and because the corpus of words piloted for each task had to be so large (n 5 99 per task), word lists of animal names, implement names, and words with emotional connotations could not be matched for frequency of usage in the English language. Frequency of usage in the English language was calculated on the basis of the norms of Francis and Kucera (1982); when a word was not represented in this corpus, it was assigned a value of zero. Average frequency of usage per million words for each category was as follows: animal mean 5 3.25, SD 5 5.69; implement mean 5 11.10, SD 5 19.90; words with emotional connotation mean 5 31.06, SD 5 61.06. However, when possible words from Toglia and Battig’s (1978) norms were selected (34 of 72 words for animal names; 37 of 72 words for implement names; 45 of 72 words with emotional connotation). All these words had imagery, concreteness, and familiarity ratings above the midpoint of the rating scale (4 on a 7-point scale), indicating that words had high imageability, concreteness, and familiarity. Using a semantic monitoring task similar to that of Binder et al. (1997, 1999) and Demonet et al. (1992), subjects monitored words for two designated criteria. For words with emotional connotations, subjects pressed a button when the item was a nonliving thing and had a negative emotional connotation (e.g., vomit and tornado would be targets; roach, puppy, and cake would be nontargets). Negative emotional connotation was selected to enhance participants’ attention to emotional attributes; the nonliving distinction was chosen so that participants would attend to the category distinction which guided the selection of semantic monitoring control tasks. Because the current study focused on a semantic attribute, emotional connotation, the semantic monitoring comparison tasks emphasized attributes which were characteristic of the categories used in these tasks and which were largely devoid of emotional connotation. For animal names, subjects pressed a button when the animal had more than two legs and was land-dwelling (e.g., moose and grasshopper would be targets; dolphin, crab, and ostrich would be nontargets). For implement names, subjects pressed a button when the implement was used primarily outdoors and required only one hand (e.g., hatchet and baseball would be targets; shovel, spatula, keyboard would be nontargets). The tasks were matched in frequency of targets (average of two targets per half cycle). Responses consisted of a thumb press to a hand-held button assembly placed in the subject’s left hand when a target was recognized. Button presses produced a visual signal in the control room. The notebook computer used for stimulus presentation recorded those items that were endorsed as targets, and measured response latency to those stimuli via a program written for this purpose. The order of the experimental runs (words with emotional connotation vs. animal names vs. implement names) was counterbalanced across subjects. B. Crosson et al. A pilot study was conducted with different subjects to equate the three experimental tasks for accuracy and reaction time. During the pilot study, a separate group of 7 subjects (5 F, 2 M; graduate students) listened to words and pressed a button in their right hand if the word met both criteria for the task. They pressed a button in their left hand if the word did not. This dual response format was used for the pilot study so that reaction time could be used as one index of the ambiguity or difficulty of semantic judgments for both target and nontarget items. Subjects performed the three monitoring tasks in separate experimental runs. Ninetynine words were used in each task, divided into 11 cycles. Words were excluded if they were excessively ambiguous or difficult, as indexed by variability in response or long reaction times. The control task was a tone monitoring task patterned after that of Binder et al. (1997). Subjects heard sequences of high- and low-frequency digitally synthesized tones. Sequences of three to seven tones were presented, and subjects pressed a button following each sequence that contained two or more high-frequency tones. Target density for tone monitoring was matched to that of the semantic monitoring tasks. The control task was selected to mitigate the effects of auditory input, attentional demands, and motor responding. We did not choose words or word-like stimuli for a control task because of their potential for activating semantic searches (see, e.g., Price et al., 1996). Image Acquisition Each experimental run consisted of 8 cycles, alternating between 28-s periods of word monitoring and 28 second periods of tone monitoring. For each functional image slice, there were 8 images per half cycle (128 images per task). Data were acquired on a 1.5 T GE Signa scanner using a dome-shaped quadrature radio frequency head coil. The head was aligned such that the interhemispheric fissure was within 18 of vertical. The most medial sagittal slice for functional images was placed such that the medial edge of the slice corresponded with the medial boundary of the left hemisphere. For f MRI sequences, nine slices (6.5 to 6.9 mm thick) were used to cover the entire left hemisphere. (At the time of the experiment, our imaging software was limited to nine slices, and the nine slices were used to cover the left hemisphere.) Functional images were acquired using a gradient echo spiral scan sequence (Noll et al., 1995) with TE 5 40 ms, TR 5 885 ms, FA 5 458, FOV 5 18 cm, 4 spirals. After functional image acquisition for all tasks, structural images were acquired of the whole brain (124 3 1.3 mm thick sagittal slices), using a 3D spoiled GRASS acquisition (TE 5 7 ms, TR 5 27 ms, FA 5 608, NEX 5 1, FOV 5 24 cm, matrix size 5 256 3 192). Image Analysis Functional images were analyzed and overlaid onto structural images using the AFNI program from the Medical Semantic monitoring of emotional connotation College of Wisconsin (Cox, 1996). To correct for head motion, all individual functional images were registered to the last functional image of the last run with an iterated differential spatial method of alignment. Subsequently, images were visually inspected for gross artifacts and viewed in cine loop to detect residual motion. If any time series of a subject was judged to contain a significant number of images with gross artifacts or residual motion, the subject’s data were removed from further analyses. To control for large vessel effects and other sources of artifact, voxels where the standard deviation of the signal was more than 5% of the mean intensity were eliminated from the functional image series for each run before further processing. Next, difference images between experimental and control tasks were created as follows. For each experimental or control task half cycle, the first three functional images were discarded because they represented transition of the hemodynamic response between the two task states. Subsequently, within each voxel, a mean signal intensity for each half cycle was derived from the remaining 5 images. For each experimental run, the control task mean was subtracted from the semantic monitoring mean for each of the 8 cycles on a voxel-by-voxel basis. To measure changes in signal intensity between comparison conditions for each subject, two types of t tests were conducted on a voxelby-voxel basis: (1) response to alternations between each semantic monitoring task (nouns with emotional connotations, animal names, and implement names) and the tonemonitoring control task were compared; (2) differences between each of the semantic monitoring tasks and the tone monitoring task were compared to differences between all other semantic monitoring tasks and the tone monitoring task. For each subject, resulting functional intensities consisted of t values for the various comparisons. Individual anatomic scans and functional images were interpolated to 1 mm 3 voxels, co-registered and converted to standard atlas space (Talairach & Tournoux, 1988). To compensate for individual subject variability in structural and functional anatomy, functional image volumes were smoothed 3 mm, by replacing each voxel value with the average value of the others in a 3-mm radius. The functional data sets were then merged across subjects by averaging the t statistics in each voxel; i.e., the functional intensity within each voxel was the average t value across the 16 subjects, hereafter referred to as tave . The procedure of averaging t values was used to control for heteroscedasticity of MR signal variation between subjects arising from sources unrelated to experimental conditions, such as differing degrees of tissue pulsatility, variability in global blood flow or hemodynamic reactivity, or scanner variability between sessions (Binder et al., 1999). Similar to the analyses of Binder et al. (1997), a probability of p # .0001 was used along with the Cornish-Fisher expansion (Fisher & Cornish, 1960) to select a threshold for rejection of the null hypothesis when semantic monitoring tasks were compared directly to the tone monitoring control task. When semantic monitoring-control task differences were compared be- 611 tween the different semantic monitoring tasks, a probability level of p # .005 (Cornish-Fisher expansion) was used as a threshold to determine significance. The relaxed probability criterion for the latter comparisons was necessary to increase sensitivity of comparisons because the semantic monitoring tasks being compared were all similar language tasks, with fewer differences in cognitive components than word monitoring versus tone monitoring comparisons. Such procedures have been used in other functional neuroimaging studies of language (e.g., Martin et al., 1996). For the comparisons among semantic monitoring tasks, all areas meeting significance criteria are listed in the results section, but interpretation focused on the a priori hypothesis that activity in the more anterior portions of the left frontal and temporal lobes would be stronger for monitoring words with emotional connotations than for monitoring animal or implement names. To further control for type I statistical error, only contiguous volumes $ 200 ml were interpreted in all comparisons. RESULTS Analysis of Behavioral Data Reaction times for items that subjects endorsed as targets were compared between word monitoring tasks to ascertain if subjects spent more time on one task versus others and as an index of task difficulty. Average reaction times for target items are presented in Table 1. A one-way repeated measures analysis of variance indicated no significant differences between word monitoring tasks in reaction times @F~2,26! 5 1.47, p 5 .25]. Thus, word monitoring tasks were equivalent with respect to this measure. Comparison of semantic monitoring tasks to tone monitoring tasks Based on previous literature, two a priori hypotheses could be evaluated in comparisons of semantic monitoring and tone monitoring tasks. The first stated that activity in the more anterior portions of the left frontal and temporal lobes would be stronger for monitoring words with emotional connotations than for monitoring animal or implement names. Second, because the semantic attributes of animals and implements include visual form, and because items with emotional connotations were selected for ease of visualization, it was hypothesized that all word monitoring tasks Table 1. Mean reaction times for target responses Task Emotional connotations Animal names Implement names Reaction time (s) SD 1.95 2.00 2.16 0.18 0.59 0.23 612 would demonstrate activity in the fusiform gyrus compared to the tone monitoring control task. Each of the semantic monitoring tasks was compared directly to the control task (monitoring of tone sequences). T values were derived within voxels for each subject and then averaged across subjects. Thus, the functional intensity for group images was an average t value, referred to as tave . A common criterion for significance was applied to all comparisons (probability for tave # .0001, cluster of contiguous activity . 200 ml). Activity in the More Anterior Portions of the Left Frontal and Temporal Lobes The first experimental hypothesis stated that semantic monitoring of words with emotional connotations would produce activity in the more anterior portions of the left frontal and temporal lobes while monitoring of animal and implement names would not produce such activity. If this hypothesis is correct, then comparison of semantic monitoring for words with emotional connotations to the tone monitoring task should evoke activity in the anterior left frontal and temporal lobes whereas experimental-control task comparisons for animal and implement names should not evoke such activity. For monitoring words with emotional connotations, Table 2 shows a large volume of activity whose maximum tave (max tave ) is in the anterior left frontal lobe (anterior superior frontal gyrus). Figure 1 indicates this activity actually begins on the crest of the left hemisphere just lateral to midline and anterior to pre-SMA. Then it extends anteriorly to cortex near the frontal pole, spanning the entirety of what Talairach and Tournoux (1988) labeled Brodmann’s areas (BAs) 8 and 9, and extending into BA10 near the frontal pole. The max tave within this cluster is somewhat posterior and superior to the activity that Crosson et al. (1999) found near the frontal pole for generation of words with emotional connotations versus emotionally neutral words (Talairach coordinates for maximum functional intensity from previous study 5 213, 62, 27). However, the activity in the current comparison subsumes the area of activity difference from the previous study. Although there was no activity at or just behind the left temporal pole in the current study, there was an area of activity in the anterior portion of the superior temporal gyrus (Table 2, Figure 2) about 20 mm behind the maximum functional intensity from the word generation study (Talairach coordinates for maximum functional intensity from the previous study 5 248, 17, 215). Thus, for monitoring of words with emotional connotations in the current study, activity in the more anterior portions of the left frontal lobe was confirmed, and anterior left temporal lobe activity was observed somewhat posterior to the temporal pole. No activity in the more anterior portions of the left frontal and temporal lobes was found for monitoring of animal names; however, monitoring of implement names produced activity in the anterior left frontal lobe and posterior to the temporal pole near the max tave for monitoring words with B. Crosson et al. emotional connotations (Figures 1 and 2, respectively). In both locations, activity was more wide spread for words with emotional connotations than for implement names. For monitoring words with emotional connotations, the activity in the anterior left frontal lobe was contiguous with activity extending posteriorly along the crest of the frontal lobe, but for implement names, activity was not contiguous along the crest of the frontal lobe. Instead, for implement names, one superior area of activity located more posteriorly in the medial frontal cortex merged with a large area of lateral frontal activity (Table 2). The spatial extent of activity in the anterior temporal lobe was about 75% greater for words with emotional connotations compared to tool names. Activity in the Fusiform Gyrus The second hypothesis predicted that in semantic monitoringtone monitoring comparisons, activity would be present in the fusiform gyrus for all three experimental conditions (monitoring words with emotional connotations, monitoring animal names, monitoring implement names) compared to the tone monitoring control task. The findings were as expected (Table 2, Figure 3). The activity was more medially located for animal names than for implement names and words with emotional connotations. For implement names, the activity merges with activity located more laterally in the inferior temporal gyrus. Other Areas of Activity For experimental–control task comparisons, a number of other volumes of significant activity were present (Table 2). All tasks share some common volumes of activity. Every semantic monitoring task evoked activity in the lateral prefrontal cortex; the volume was largest and quite extensive for monitoring of implement names, and it was smallest but still substantial for animal names. Monitoring of all three word types also elicited activity centering in the angular gyrus; again, the activity was most extensive and occupied surrounding areas for implements, and it was least extensive for animals. In addition to activity in the more anterior portion of the left frontal lobe and activity in the anterior temporal cortex, monitoring of words with emotional connotations and implement names had small volumes of activity in common for two other regions. Compared to the tone monitoring control task, these two experimental tasks both evoked a separate volume of activity in pre-SMA; the activity was slightly more superior in monitoring of emotional words. The posterior cingulate region also demonstrated activity during both tasks, though the activity for monitoring of words with emotional connotations was anterior and superior to that for monitoring implement names. Monitoring of implement names and animal names also showed activity in common for three other regions. First, compared to the tone monitoring task, both of these semantic monitoring tasks evoked activity in the superior portion Semantic monitoring of emotional connotation 613 Table 2. Volumes of tissue showing significant activity changes for semantic monitoring tasks versus tone monitoring task Volumes of activity change . 200 ml, tave p , .0001 Location Monitor emotional connotation vs. monitor tone sequences Monitor tool names vs. monitor tone sequences Monitor animal names vs. monitor tone sequences ml anatomic area (max tave loc) max tave ml anatomic area (max tave loc) max tave ml anatomic area (max tave loc) max tave Semantic Monitoring > Tone Monitoring Cortex in anterior frontal lobe 8815 BA 8, 9, 10 (212,50,46) max tave 5 3.16 608 BA 10 (211,66,21) max tave 5 1.77 Anterior temporal cortex 859 BA 21, 22 (258,23,211) 489 BA 21, 22 (257,22,25) max tave 5 1.68 max tave 5 1.68 Fusiform gyrus 275 BA 37 (242,247,214) max tave 5 1.63 4053 BA 37 (243,242,212) max tave 5 2.13 224 BA 37 (234,233,220) max tave 5 1.61 6061 BA 9, 10, 44, 45, 47 (252,26,15) max tave 5 2.22 22155 BA 8, 9, 10, 44 45, 46, 47 (249,44,3) max tave 5 3.24 1518 BA 9, 10, 45 (250,32,19) max tave 5 2.41 Other frontal activity Lateral Frontal Pre-SMA0other medial frontal 351 BA 6 (25,23,59) max tave 5 1.82 349 BA 6 (25,21,48) max tave 5 1.62 (Second volume extends from lateral frontal volume across crest of frontal lobe, BAs 6, 8, 9) Superior premotor Other inferior temporal Angular gyrus 3236 BA 39 (249,261,26) max tave 5 2.72 Occipitoparietal Posterior cingulate 527 BA 6, 8 (235,17,42) max tave 5 1.74 1106 BA 6, 8 (226,20,53) max tave 5 2.17 (Extends from peak in fusiform gyrus to inferior temporal gyrus, BA 37) 1101 BA 37 (248,257,210) max tave 5 1.97 7619 BA 19, 39, 40 (249,264,31) 603 BA 39 (250,261,26) max tave 5 2.83 max tave 5 1.89 (Extends from peak in angular gyrus to occipital cortex, BA 19) 397 BA 23 (24,242,27) max tave 5 1.76 257 BA 19 (233,266,40) max tave 5 1.70 316 BA 30 (28,254,11) max tave 5 1.64 Tone Monitoring > Semantic Monitoring Other frontal SMA 308 BA 6(24,21,62) max tave 5 21.52 Motor Superior Premotor Temporal 324 BA 4 (257, 1, 17) max tave 5 21.99 309 BA 4, 44 (256, 7, 6) max tave 5 21.58 1036 BA 6, 4 (242, 26, 57) max tave 5 22.04 839 BA 6, 4 (242, 26, 57) max tave 5 21.99 1106 BA 6, 4 (242, 27, 57) max tave 5 2.17 412 BA 42, 22 (263,235,19) max tave 5 21.69 702 BA 42, 22 (264,233,19) max tave 5 21.82 269 BA 41, 42, 22 (250, 26, 5) max tave 5 21.75 725 BA 42, 22 (263,233,20) max tave 5 2.24 BA 5 putative Brodmann’s Area (according to Talairach & Tournoux, 1988); max tave 5 maximum tave within given cluster of activity. 614 B. Crosson et al. Fig. 1. Significant activity ( p , .0001) along crest of the left frontal lobe, including cortex near frontal pole, for emotional connotation vs. tones (upper left) and for implement names vs. tones (lower right). Significant activity is in black. No significant activity was present for animal names vs. tones (not shown). Anatomical underlay is averaged across 16 subjects. Left two images in each montage are sagittal planes, and right two images are coronal planes. L 5 distance in mm to left of midline; A 5 distance in mm anterior to anterior commissure. of premotor cortex. Second, activity extended from its maximum functional intensity in the fusiform gyrus into the inferior temporal gyrus for monitoring of implement names; activity in the inferior temporal gyrus was a separate volume of activity for monitoring of animal names. Third, activity extended from its maximum functional intensity in the angular gyrus to occipital cortex (BA 19) for monitor- ing of implement names; activity in this latter area was a separate volume for monitoring of animal names. The tone monitoring task also evoked a few areas of significant activity. First, compared to each semantic monitoring task, the tone monitoring task produced a significant activity cluster in the superior temporal gyrus close to primary auditory cortex. Second, compared to each semantic Semantic monitoring of emotional connotation 615 Fig. 2. Significant activity ( p , .0001) in anterior temporal lobe for emotional connotation vs. tones (upper left) and for implement names vs. tones (lower right). Significant activity is in black. No significant activity was present for animal names vs. tones (not shown). Anatomical underlay is averaged across 16 subjects. Left two images in each montage are sagittal planes; right two images are coronal planes. L 5 distance in mm to left of midline; P 5 distance in mm posterior to anterior commissure monitoring task, the tone monitoring task produced a significant cluster in superior premotor cortex. For both the superior temporal and the superior premotor clusters, the local maxima are in almost exactly the same location across comparisons with all three semantic monitoring tasks. The tone monitoring task produced two unique areas of activity in motor cortex when compared to the implement monitoring tasks, and one unique area of activity in SMA when compared to monitoring of words with emotional connotations. Comparison of Monitoring Words with Emotional Connotations to Other Experimental Tasks Comparisons among semantic monitoring tasks also were made by contrasting semantic monitoring minus tone monitoring signal intensity differences from one word monitoring task to the signal intensity differences from another semantic monitoring minus tone monitoring comparison. 616 B. Crosson et al. Fig. 3. Significant activity ( p , .0001) in the fusiform gyrus (posterior inferior temporal lobe) for emotional connotation vs. tones (upper right), for animal names vs. tones (lower left), and for implement names vs. tones (lower right). Black arrows point to this activity, which is represented in black. Anatomical underlay is averaged across 16 subjects. Left two images in each montage are sagittal planes; right two images are coronal planes. L 5 distance in mm to left of midline; P 5 distance in mm posterior to anterior commissure. Similar to other analyses, tave was computed by averaging t values from individual subject images. As in similar studies (e.g., Martin et al., 1996), statistical criteria were relaxed somewhat (probability for average t # .005, volume of activity . 200 ml) to increase sensitivity for comparisons that are more closely matched in their cognitive components. All clusters meeting these criteria are listed in Table 3. Since the criterion was relaxed, we focus primarily on the experimental hypotheses that monitoring of words with emotional connotations more intensely involves cortex in the anterior portions of the left frontal and temporal lobes than monitoring of animal or implement names. When monitoring of words with emotional connotations was compared to monitoring of implement names in this fashion, two areas of activity arose in the anterior left frontal lobe (Figure 4). The maximum intensity of the volume closest to the frontal pole (Talairach coordinates for max tave 5 213, 63, 29) was within 2 mm or less in every plane to the maximum intensity of a similar volume found in comparison of generation of words with emotional connotations to generation of emotionally neutral words (Crosson et al., 1999). The second anterior frontal volume for this comparison was somewhat posterior and superior to the first. When monitoring of words with emotional connotations was compared to monitoring of animal names, one volume of activity in the more anterior portion of the left frontal lobe emerged (Figure 4); this volume of activity subsumed both activity volumes from the emotional connotation versus implement name comparison. The fact that activity in the more anterior aspect of the left frontal lobe was greater for monitoring words with emotional connotation than for monitoring either animal names or implement names is evident from the average time series plotted in Figure 4. Monitoring of words with emotional connotations did not produce a significant volume of activity in the anterior temporal lobe compared either to monitoring of implement names or to monitoring of animal names. Semantic monitoring of emotional connotation 617 Table 3. Volumes of activity change . 200 ml, tave p , .005 Location Cortex near frontal pole Monitor emotional connotation vs. monitor implement names Monitor emotional connotation vs. monitor animal names Monitor implement names vs. monitor animal names ml anatomic area (max tave loc) max tave ml anatomic area (max tave loc) max tave ml anatomic area (max tave loc) max tave Emotional connotation > animal names 3019 BA 8, 9, 10 (211, 50, 46) max tave 5 1.70 Implement names > animal names Emotional connotation > implement names 842 BA 8, 9 (28, 53, 45) max tave 5 1.22 268 BA 9, 10 (213, 63, 29) max tave 5 1.30 Other frontal 205 BA 47 (250, 34, 24) max tave 5 1.04 Lateral frontal Premotor 488 BA 47 (252, 33, 21) max tave 5 1.24 1430 BA 6, 8 (219, 25, 53) max tave 5 1.31 Implement name > emotional connotation Animal names > emotional connotation Animal names > implement names Other Frontal Lateral frontal 237 BA 8, 9 (243, 31, 28) max tave 5 21.09 Premotor 2245 BA 6 (220, 11, 52) max tave 5 21.25 751 BA 6, 8 (249, 8, 29) max tave 5 21.16 Temporal 1460 BA 21, 37 (259, 251, 23) max tave 5 21.59 Parietal occipital0parietal 1552 BA 40, 19 (234, 278, 36) max tave 5 21.62 365 BA 21, 37 (253, 252, 21) max tave 5 21.09 206 BA 7 (26, 268, 48) max tave 5 21.26 Insula 589 Insula (240, 240, 42) max tave 5 21.13 BA 5 putative Brodmann’s Area (according to Talairach & Tournoux, 1988); max tave 5 maximum tave within given cluster of activity. One other region of activity is worth noting. Monitoring of implement names produced a volume of significant activity in the superior premotor cortex compared both to monitoring of animal names and to monitoring of words with emotional connotations (Figure 4). The fact that activity in the superior premotor cortex was greater for monitoring implement names than for monitoring either animal names or words with emotional connotations also can be discerned from the average time series plotted in Figure 4. This volume is important because it demonstrates that activity along the crest of the left frontal lobe follows an anterior to posterior gradient for monitoring of words with emotional connotations versus monitoring of implement names. Although both tasks demonstrate activity in both regions compared to tone monitoring, monitoring of words with emotional connotations evoked greater activity near the frontal pole while monitoring of implement names evoked greater activity in superior premotor cortex. DISCUSSION This study was designed as a follow-up to our earlier study on generation of words with emotional connotations (Crosson et al., 1999) and was based on the theory that a phonological input lexicon, a phonological output lexicon, and a structural (visual) description system all have separate neural representations which must access semantic information to assign meaning to these representations (e.g., Ellis &Young, 1988). Under this theory, a neural mechanism involved in processing the semantic attribute of emotional connotation would be active whether the task involved input through the phonological input lexicon or the structural description 618 B. Crosson et al. Fig. 4. Sagittal and coronal images in the upper left show significant activity in black ( p , .005) near the left frontal pole for emotional connotation minus tones . implement names minus tones (top) and for emotional connotation minus tones . animal names minus tones (bottom). Anatomical underlay is averaged across the 16 subjects. The time series in the upper right were derived from the large area near the frontal pole in the emotional connotation minus tones . implement names minus tones comparison. A mask of this area was created, voxels within the mask were averaged into a single time series for each individual participant, the resulting time series was subjected to slight temporal smoothing, and this single average time series was averaged across participants. Sagittal and coronal images in the lower right demonstrate activity in white ( p , .005) in the left superior premotor cortex for implement names minus tones . emotional connotation minus tones (top) and for implement names minus tones . animal names minus tones (bottom). The time series in the lower left were derived from the area in superior premotor cortex in the implement names minus tones . emotional connotation minus tones comparison. For all time series graphs in this figure, the x axis represents eight 56-s cycles alternating between semantic monitoring and tone monitoring, and the y axis is scaled to represent a 1.4% change in signal intensity from the bottom to the top of the graph. The asterisks indicate the time series for which semantic monitoring minus tone monitoring differences are significantly greater than for the other tasks. L 5 distance in mm to the left of midline; A 5 distance in mm anterior to anterior commissure. system or involved accessing the phonological output lexicon. Anterior left frontal cortex previously has been shown to be active when processing pictures with emotional content (Lane et al., 1997; Paradiso et al., 1999), and when accessing the phonological output lexicon during production of words with emotional connotation (Crosson et al., 1999). Anterior left temporal cortex previously has shown activity when processing pictures with positive emotional content (Lane et al., 1999), and when accessing the phonological output lexicon during production of words with emotional connotation (Crosson et al., 1999). Therefore, if words with emotional connotations presented via the phonological input lexicon activated anterior left frontal and temporal cortices in comparison to emotionally neutral words, the inference that these areas are involved in processing the semantic attribute of emotional connotation would be strengthened. Semantic monitoring of emotional connotation Thus, this study was designed to ascertain if processing of words with emotional connotations, presented via the auditory modality, would evoke activity in the more anterior portions of the left frontal and temporal lobes, thereby confirming the involvement of these cortices in semantic processing of emotional connotation. Further, emotional connotation was hypothesized to be a semantic attribute that cuts across traditional semantic categories. For example, both living and nonlivings things may have emotional connotations (e.g., roach and coffin, respectively). To assess this aspect of emotional connotation as a semantic attribute, words with emotional connotations were chosen from both living and nonliving categories, and semantic monitoring of these words was compared to semantic monitoring of emotionally neutral words from two categories, animal names and implement names. Cortex near the left frontal pole was active when processing words with emotional connotations was compared to the tone monitoring control task, monitoring of implement names, or monitoring of animal names, suggesting it is involved in semantic processing of emotional connotation. Cortex in the anterior temporal lobe was active only when monitoring of words with emotional connotations was compared to the tone monitoring control task, leaving some doubt as to whether it is involved in processing emotional connotation as a semantic attribute, or whether some other factor accounts for its activity in our previous study (Crosson, et al., 1999). An alternative strategy to assess involvement of the anterior left frontal cortex in processing the semantic attribute of emotional connotation might have been to give subjects words with emotional connotations in both a semantic monitoring task and a phonological monitoring task. If anterior left frontal cortex were active during the semantic but not the phonological task, then it would be involved in semantic processing. Such semantic processing could be shown to be exclusive for emotional connotation if words from other categories were subjected to similar comparisons and did not activate the cortex in question. We chose not to use such a strategy for two reasons: First, previous work by Price et al. (1996) has shown that processing of real words activates areas not activated by processing pronounceable nonwords when words and nonwords are processed in an identical task involving neither lexical nor semantic features. Thus, it may be difficult not to process lexical– semantic features of real words, even when tasks do not require such processing. This may be particularly true for words with emotional connotations because of the salience of this property, and would weaken or negate the ability to image cortex involved in processing emotional connotation. Second, in the current study, the direct comparison of semantic monitoring of words with emotional connotation to semantic monitoring of animal and implement names was used to control for areas involved in more general aspects of semantic processing. Limitations of the current study should be mentioned. For example, the characteristics that were chosen for each semantic monitoring task were carefully selected to facili- 619 tate the comparison of monitoring words with emotional connotations to monitoring animal and tool names. However, within the design of the current study, it can be argued that the type of word being monitored (i.e., words with emotional connotations vs. words from the categories of animals or implements) cannot be separated from the characteristics being monitored (i.e., “evokes a negative response and is nonliving” vs. “has more than two legs and is land dwelling” or “is used with one hand and is used outdoors”). However, when this study is considered in the context of the previous studies of word generation (Crosson et al., 1999) and viewing pictures (Lane et al., 1997; Paradiso et al., 1999), a strong case can be built for emotional connotation of words as the semantic attribute responsible for activity in the more anterior portion of the left frontal lobe. Although type of word and characteristic being monitored cannot be separated in the current study, no such problem existed in the previous study of word generation (Crosson et al., 1999). In that study, subjects simply generated words from a given category; some categories were chosen to evoke words with emotional connotations, and some were chosen to evoke emotionally neutral words. Cortex in the more anterior portion of the left frontal lobe also has been active during viewing of pictures with emotional content (Lane et al., 1997; Paradiso et al., 1999), and Beauregard et al. (1997) showed left anterior frontal activation for passive viewing of words with emotional connotation versus concrete words, even though no active processing of words was required. Activity in the latter study was somewhat posterior to activity in the current study. Thus, since no problem separating monitored characteristics from semantic attribute existed in these previous studies, we must conclude that processing the attribute of emotional connotation is responsible for evoking activity in the anterior left frontal lobe in these studies. The lack of anterior left frontal activity in Maddock and Buonocore’s (1997) work probably relate to methodological differences. Subjects passively listened to words, as opposed to actively processing them words, and activity only from threat related words was compared to activity from emotionally neutral words. A second limitation of the current study involves our decision to use the nine slices available in our image acquisition sequence to cover the entire left hemisphere. While this decision was based upon evidence of left-hemisphere dominance for processing words with or without emotional connotations (e.g., Ali & Cimino, 1997; Heilman et al., 1993), the lack of information regarding right-hemisphere activity precludes examining the role of the right hemisphere in processing words with emotional connotation. Some data suggest the right hemisphere is dominant for processing emotions (e.g., Gainotti, 1997; Heilman et al., 1984; Ross, 1997), and other data suggest the right hemisphere plays a role in processing negative (as opposed to positive) emotions (e.g., Ali & Cimino, 1997; Canli et al., 1998). However, previous research has shown that making linguistic inferences regarding emotional connotation is not impaired in patients with right-hemisphere lesions (e.g., 620 Blonder et al., 1991). Further, Cato (2001) recently conducted an f MRI study in which subjects generated words with positive emotional connotations, words with negative emotional connotations, and emotionally neutral words. In comparison to generating emotionally neutral words, generating words with positive or negative emotional connotations engaged primarily left-hemisphere mechanisms. Only for anterior frontal cortex and retrosplenial cortex did activity extend into the right hemisphere, and for both positive and negative emotional connotations, activity was far more extensive in the left than the right hemisphere for both areas. Thus, Cato’s (2001) findings indicate that normal participants rely primarily on left-hemisphere mechanisms for processing the emotional connotation of words, and that the current findings regarding left-hemisphere activity during monitoring of emotional connotation would not be greatly changed by information regarding righthemisphere activity. A feature in both our previous study (Crosson et al., 1999) and the current study was that words with positive and negative emotional connotation were grouped together. As just noted, the findings of Cato (2001) found similar patterns of activity in anterior left frontal cortex for generating words with both positive and negative emotional connotation. Thus, cortex near the left frontal pole appears to be involved in processing of emotional connotation of words without respect to valence. Additional areas might show activity if processing of positive and negative connotations were segregated in a different experimental design. It also is possible that emphasis on monitoring for negative emotion in the current study weakened activity in the anterior left temporal lobe compared to our previous study of word generation (Crosson et al., 1999) since the left hemisphere is thought to be more involved in processing positive than negative emotion (Canli et al., 1998; Heilman et al., 1993). Our secondary hypothesis that all semantic monitoringtone monitoring comparisons would demonstrate activity in the fusiform gyrus also was confirmed. This finding reflects the importance of visual form to semantic processing of words in our tasks. Visual form has long been presumed an important distinction in semantic processing of animals. While implements are defined by function, they also are recognized by visual form. Since words with emotional connotations were chosen for ease of visualization, visual form was important in their identification as well. Activity in the ventral visual stream of our subjects is consistent with recent findings demonstrating activity in the fusiform gyrus during processing of animals and tools, whether during picture naming, semantic monitoring of words, or visual tasks (Chao et al., 1999; Damasio et al., 1996; Ishai et al., 1999; Martin et al., 1996). Similarities between monitoring words with emotional connotations and monitoring implement names should be noted. Compared to tone monitoring, monitoring of both words with emotional connotations and implement names evoked activity along the crest of the left frontal lobe, including cortex in the left anterior frontal lobe, and activity B. Crosson et al. in the anterior temporal lobe. Lesion data of Lu et al. (2002) and of Cappa et al. (1998) confirm left anterior temporal involvement in naming implements as opposed to living things. Although activity along the crest of the left frontal lobe occurred for both conditions relative to tone monitoring, it should be noted that there was an anterior to posterior gradient differentiating the two tasks (Figure 4). In cortex near the left frontal pole, which shows greater connectivity to limbic regions, activity for monitoring words with emotional connotations was significantly stronger than activity for monitoring implement names. Greater involvement of this cortex in processing emotional connotations is consistent with access to information from the limbic system. In particular, cortex in the more anterior portions of the frontal lobe, including Brodmann’s area 9, is connected with paralimbic association areas, including the anterior cingulate gyrus, the posterior cingulate gyrus, retrosplenial cortex, and the posterior parahippocampal gyrus (Pandya & Yeterian, 1985). In superior premotor cortex, however, activity for monitoring implement names was significantly stronger than activity for monitoring words with emotional connotations. Greater involvement of this cortex in processing implement names is consistent with the fact that information about movement sequences is central to understanding the nature of specific implements. Our data indicate that some cortical regions supporting lexical–semantic processing cut across semantic category and item attributes. An area around the angular gyrus and a lateral frontal area encompassing Broca’s area were active in all semantic monitoring-tone monitoring comparisons. Binder et al. (1999) recently demonstrated that the angular gyrus and an area in the dorsal frontal lobe demonstrated more activity during semantic monitoring (animal names) than phoneme monitoring. The angular gyrus activity in the current study overlapped with that of Binder et al.; however, the dorsal frontal activity of Binder et al. was generally superior to the common activity of the current study. Anomic aphasia often occurs in lesions of the angular gyrus (Goodglass & Kaplan, 1983), and lesions in this general region also may cause transcortical sensory aphasia (Alexander, 1997). Anomic aphasia also may be found in lesions extending into Broca’s area from the cortex just anterior to it; transcortical motor aphasia also may occur as a result of lesions in this region (Alexander, 1997). Anomic aphasia appears to be due to an inability to link semantic information to the corresponding lexical items during language output. Semantic components also are thought to exist for transcortical sensory aphasia (Alexander, 1997) and transcortical motor aphasia (McCarthy & Warrington, 1984). Thus, when examined in light of existing literature, our data suggest areas in the left angular gyrus and left frontal lobe play a role in coordinating semantic functions for language, perhaps linking lexical and semantic retrieval processes. A brief discussion of the difference in monitoring animal names versus implement names is in order. The major difference in direct comparison of these tasks was the greater activity in the premotor cortex for monitoring implement Semantic monitoring of emotional connotation names than for monitoring animal names. The more inferior of our premotor activations for this comparison was consistent with Martin et al.’s (1996) finding in a comparison of animal and tool naming. The fact that premotor cortex is important in processing implement names lends credence to the notion that knowledge about the movements involved in using implements is a central semantic feature of these items. Indeed, Damasio and Tranel (1993) found that lesions in this area affected action naming. In conclusion, we return to the concept of emotional connotation as a semantic attribute. Findings from the current study, especially in the context of previous studies (Beauregard et al., 1997; Crosson et al., 1999; Lane et al., 1997; Paradiso et al., 1999), provide strong evidence that cortex near the left frontal pole is involved in processing emotional connotation as a semantic attribute. Emotional connotation appears to cut across the traditional semantic categories of living versus nonliving things and is largely orthogonal to that distinction. In other words, cortex near the left frontal pole appears to be important for processing emotional connotation without respect to whether the item being processed is a living or a nonliving thing. This result is important not only for understanding how emotional connotation is processed, it is important because it illustrates how attributes in general can be a dimension of semantic processing that is orthogonal to semantic category. Although past studies (Mummery et al., 1998; ThompsonSchill et al., 1999) have addressed the issue of semantic attributes, they have confounded the living–nonliving distinction with attributes processed in the ventral visual stream. Since processing of living things is thought to be dependent on processing in the ventral visual stream, these studies have not been able to demonstrate that processing of a semantic attribute may be independent from processing categories. Future research should continue to focus on distinctions between processing of semantic category and semantic attribute. ACKNOWLEDGMENTS Supported by McKnight Brain Institute of the University of Florida and Grant # R01-DC03455 from NIDCD. Thanks to Jeffrey R. Binder (Grant # R01 NS33576 from NINDS) for providing the semantic and tone monitoring tasks after which our tasks were modeled. Thanks also to NIMH Center for the Study of Emotion and Attention (Grant # P50-MH52384, P.J. Lang, Director; Margaret M. Bradley; Bruce N. Cuthbert) for consultation; Douglas C. Noll for spiral scan sequences; Jeffrey R. Fitzsimmons for head coil; Wayne King for sound transducer. REFERENCES Alexander, M.P. (1997). Aphasia: Clinical and anatomic aspects. In T.E. Feinberg & M.J. Farah (Eds.), Behavioral neurology and neuropsychology (pp. 133–149). New York: McGraw-Hill. Ali, N. & Cimino, C.R. (1997). Hemispheric lateralization of perception and memory for emotional verbal stimuli in normal individuals. Neuropsychology, 11, 114–125. 621 Beauregard, M., Chertkow, H., Bub, D., Murtha, S., Dixon, R., & Evans, A. (1997). The neural substrate for concrete, abstract, and emotional word lexica: A positron emission tomography study. Journal of Cognitive Neuroscience, 9, 441– 461. Binder, J.R., Frost, J.A., Hammeke, T.A., Cox, R.W., Rao, S.M., & Prieto, T. (1997). Human brain language areas identified by functional magnetic resonance imaging. Journal of Neuroscience, 17, 353–362. Binder, J.R., Frost, J.A., Hammeke, T.A., Bellgowan, P.S.F., Rao, S.M., & Cox, R.W. (1999). Conceptual processing during the conscious resting state: A functional MRI study. Journal of Cognitive Neuroscience, 11, 80–93. Blonder, L.X., Bowers, D., & Heilman, K.M. (1991). The role of the right hemisphere in emotional communication. Brain, 114, 1115–1127. Bradley, M.M., Cuthbert, B.N., & Lang, P.J. (1988). Affective Norms for English Words (ANEW). Technical manual and affective ratings. Gainesville, FL: University of Florida Center for Research in Psychophysiology. Canli, T., Desmond, J.E., Zhao, Z., Glover, G., & Gabrieli, J.D.E. (1998). Hemispheric asymmetry for emotional stimuli detected with f MRI. NeuroReport, 9, 3233–3239. Cappa, S.F., Frugoni, M., Pasquali, P., Perani, D., & Zorat, P. (1998). Category-specific naming impairment for artefacts: A new case. NeuroCase, 4, 391–397. Caramazza, A. & Shelton, J.R. (1998). Domain-specific knowledge systems in the brain: The animate–inanimate distinction. Journal of Cognitive Neuroscience, 10, 1–34. Cato, M.A. (2001). Emotional connotation as a semantic attribute: A whole brain fMRI study of emotional valence and sex effects on word generation. Unpublished doctoral dissertation, University of Florida, Gainesville. Chao, L.L., Haxby, J.V., & Martin, A. (1999). Attribute-based neural substrates in temporal cortex for perceiving and knowing about objects. Nature Neuroscience, 2, 913–919. Cox, R.W. (1996). AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages. Computers in Biomedical Research, 29, 162–173. Crosson, B., Radonovich, K., Sadek, J.R., Gökçay, D., Bauer, R.M., Fischler, I.S., Cato, M.A., Maron, L., Auerbach, E.J., Browd, S.R., & Briggs, R.W. (1999). Left-hemisphere processing of emotional connotation during word generation. NeuroReport, 10, 2449–2455. Damasio, H., Grabowski, T.J., Tranel, D., Ponto, L.L.B., Hichwa, R.D., & Damasio, A.R. (1999). Neural correlates of retrieval of words for spatial relationships. NeuroImage, 9, S918. Damasio, A.R. & Tranel, D. (1993). Nouns and verbs are retrieved with differently distributed neural systems. Proceedings of the National Academy of Science, USA, 90, 4957– 4960. Demonet, J.-F., Chollet, R., Ramsay, S., Cardeat, D., Nespoulous, J.L., Wise, R., Rascol, A., & Frackowiak, R.S.J. (1992). The anatomy of phonological and semantic processing in normal subjects. Brain, 115, 1753–1768. Ellis, A.W. & Young, A.W. (1988). Human cognitive neuropsychology. London: Erlbaum. Ferreira, C.T., Fiusiano, B., & Poncet, M. (1997). Categoryspecific anomia: Implication of different neural networks in naming. NeuroReport, 8, 1595–1602. Fisher, R.A. & Cornish, E.A. (1960). The percentile points of distributions that have known cumulants. Technometrics, 2, 209–226. 622 Francis, W.N. & Kucera, H. (1982). Frequency analysis of English usage: Lexicon and grammar. Boston: Houghton Mifflin. Gainotti, G. (1997). Emotional disorders in relation to unilateral brain damage. In T.E. Feinberg & M.J. Farah (Eds.), Behavioral neurology and neuropsychology (pp. 691– 698). New York: McGraw-Hill. Goodglass, H. & Kaplan, E. (1983). The assessment of aphasia and related disorders. Philadelphia: Lea & Febiger. Hart, J. & Gordon, B. (1992). Neural subsystems for object knowledge. Nature, 359, 60– 64. Heilman, K.M., Bowers, D., Speedie, L., & Coslett, H.B. (1984). Comprehension of affective and nonaffective prosody. Neurology, 34, 917–921. Heilman, K.M., Bowers, D., & Valenstein, E. (1993). Emotional disorders associated with neurological diseases. In K.M. Heilman & E. Valenstein (Eds.), Clinical neuropsychology (3rd ed., pp. 461– 497). New York: Oxford University Press. Hillis, A.E. & Caramazza, A. (1991). Category-specific naming and comprehension impairment: A double dissociation. Brain, 114, 2081–2094. Ishai, A., Ungerleider, L.G., Martin, A., Schouten, J.L., & Haxby, J.V. (1999). Distributed representation of objects in the human ventral visual pathway. Proceedings of the National Academy of Science, USA, 96, 9379–9384. Lane, R.D., Reiman, E.M., Bradley, M.M., Lang, P.J., Ahern, G.L., Davidson, R.J., & Schwartz, G.E. (1997). Neuroanatomic correlates of pleasant and unpleasant emotion. Neuropsychologia, 35, 1437–1444. Lane, R.D., Chua, P.M.-L., & Dolan, R.J. (1999). Common effects of emotional valence, arousal and attention on neural activation during visual processing of pictures. Neuropsychologia, 37, 989–997. Lu, L.H., Crosson, B., Nadeau, S.E., Heilman, K.M., Rothi, L.J.G., Raymer, A., Gilmore, R.L., Bauer, R.M., & Roper, S.N. (2002). Category-specific naming deficits for objects and actions: Semantic attribute and grammatical role hypotheses. Neuropsychologia, 40, 1608–1621. Maddock, R.J. & Buonocore, M.H. (1997). Activation of left posterior cingulate gyrus by the auditory presentation of threat-related words: An f MRI study. Psychiatry Research, 75, 1–14. Martin, A., Wiggs, C.L., Ungerleider, L.G., & Haxby, J.V. (1996). Neural correlates of category-specific knowledge. Nature, 379, 649– 652. McCarthy, R.A. & Warrington, E.K. (1984). A two-route model of speech production: Evidence from aphasia. Brain, 107, 463– 486. B. Crosson et al. McCarthy, R.A. & Warrington, E.K. (1988). Evidence for modalityspecific meaning systems in the brain. Nature, 344, 428– 430. Mummery, C.J., Patterson, K., Hodges, J.R., & Price, C.J. (1998). Functional neuroanatomy of the semantic system: Divisible by what? Journal of Cognitive Neuroscience, 10, 766–786. Noll, D.C., Cohen, J.D., Meyer, C.H., & Schneider, W.J. (1995). Spiral k-space MR imaging of cortical activation. Magnetic Resonance Imaging, 5, 49–56. Oldfield, R.C. (1971). The assessment and analysis of handedness: The Edinburgh Inventory. Neuropsychologia, 9, 97–113. Pandya, D.N. & Yeterian, E.H. (1985). Architecture and connections of cortical association areas. In A. Peters & E.G. Jones (Eds.), Cerebral cortex. Vol. 4: Association and auditory cortices (pp. 3– 61). New York: Plenum Press. Paradiso, S., Johnson, D.L., Andreasen, N.C., O’Leary, D.S., Watkins, G.L., Boles Ponto, L.L., & Hichwa, R.D. (1999). Cerebral blood flow changes associated with attribution of emotional valence to pleasant, unpleasant, and neutral visual stimuli in a PET study of normal subjects. American Journal of Psychiatry, 156, 1618–1629. Price, C.J., Wise, R.J.S., & Frackowiak, R.S.J. (1996). Demonstrating the implicit processing of visually presented words and pseudowords. Cerebral Cortex, 6, 62–70. Ross, E.D. (1997). The aprosodias. In T.E. Feinberg & M.J. Farah (Eds.), Behavioral neurology and neuropsychology (pp. 699– 710). New York: McGraw-Hill. Sacchett, C. & Humphreys, G.W. (1992). Calling a squirrel a squirrel but a canoe a wigwam: A category-specific deficit for artefactual objects and body parts. Cognitive Neuropsychology, 9, 73–86. Talairach, J. & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain. New York: Thieme Medical Publishers. Thompson-Schill, S.L., Aguirre, G.K., D’Esposito, M., & Farah, M.J. (1999). A neural basis for category and modality specificity of semantic knowledge. Neuropsychologia, 37, 671– 676. Toglia, M.P. & Battig, W.F. (1978). Handbook of semantic word norms. Hillsdale, NJ: Lawrence Erlbaum Associates. Tranel, D., Damasio, H., & Damasio, A.R. (1997). A neural basis for the retrieval of conceptual knowledge. Neuropsychologia, 35, 1319–1327. Ungerleider, L.G. & Mishkin, M. (1982). Two cortical visual systems. In D.J. Ingle, M.A. Goodale, & R.J.W. Mansfield (Eds.), The analysis of visual behavior (pp. 549–586). Cambridge, MA: MIT Press. Warrington, E.K. & McCarthy, R. (1983). Category specific access dysphasia. Brain, 106, 859–878. Warrington, E.K. & Shallice, T. (1984). Category specific semantic impairments. Brain, 107, 829–854.