|Home | About | Journals | Submit | Contact Us | Français|
This study examined how task (implicit versus explicit) and semantic relationship (direct versus indirect) modulated hemodynamic activity during lexico-semantic processing. Participants viewed directly related, indirectly related and unrelated prime-target word-pairs as they performed (a) an implicit Lexical Decision (LD) task in which they decided whether each target was a real word or a nonword, and (b) an explicit Relatedness Judgment (RJ) task in which they determined whether each word-pair was related or unrelated in meaning. Task influenced both the polarity and neuroanatomical localization of hemodynamic modulation. Semantic relationship influenced the neuroanatomical localization of hemodynamic modulation. The implicit LD task was primarily associated with inferior prefrontal and ventral inferior temporal/fusiform hemodynamic response suppression to directly related (relative to unrelated) word-pairs, and with more widespread temporal-occipital response suppression to indirectly related (relative to unrelated) word-pairs. In contrast, the explicit RJ task was primarily associated with left inferior parietal hemodynamic response enhancement to both directly and indirectly related (relative to unrelated) word-pairs, as well as with left inferior prefrontal hemodynamic response enhancement to indirectly related (relative to unrelated) word-pairs. These findings are discussed in relation to the specific neurocognitive processes thought to underlie implicit and explicit semantic processes.
It has long been recognized that words and concepts that have been frequently encountered together and/or that share perceptual or functional features are structured and organized according to such associations and common features within semantic memory. There have been two main approaches to understanding how this organized structure impacts upon processing new incoming words: the first is to present participants with word-pairs or lists but not to alert them to the presence of any semantic relationship between these words, and to examine the effects of any semantic relationship between them on their performance of an incidental implicit task. The second is to alert participants to the existence of potential semantic relationships between word-pairs or lists and to ask them to use such relationships to perform a more explicit task. The current study used functional magnetic resonance imaging (fMRI) to examine the neural basis of implicit and explicit semantic processing of the same words in the same participants. We tested the overall hypothesis that both the task performed by subjects (implicit versus explicit) and the nature of the semantic relationship between pairs of words (directly related, indirectly related and unrelated) would influence both the neuroanatomical localization and the polarity of hemodynamic modulation. Understanding these relationships between task, semantic relationship and brain activity may give new insights into whether distinct semantic cognitive processes are mediated by distinct neuroanatomical networks. This, in turn, may help explain the cognitive basis of abnormal patterns of hemodynamic activity observed during semantic processing in patients with neuropsychiatric disorders including schizophrenia (Kuperberg, et al. 2007) and Alzheimer’s disease (Grossman, et al. 2003).
The most common paradigm used to explore implicit semantic processing is the semantic priming paradigm in conjunction with an implicit task. The semantic priming effect describes the shorter reaction times (Meyer and Schvaneveldt 1971; Neely 1991) or the attenuated electrophysiological response (the N400 event-related potential) (Bentin, et al. 1985; Rugg 1985) to target words (e.g. “tiger”) that are preceded by semantically related prime words (e.g. “lion”) relative to those that are preceded by semantically unrelated prime words (e.g. “truck”). One implicit task often used to study semantic priming is Lexical Decision (LD). In this task, letter-strings are randomly introduced as targets to unrelated primes, and participants are simply asked to decide whether or not each target that they encounter is a real word or a nonword (Meyer and Schvaneveldt 1971). The semantic priming effect observed during LD can reflect the operation of automatic processes such as a spread of activation across semantic memory (Anderson 1983; Collins and Loftus 1975), and/or controlled processes such as the generation of expectancies of which word will be presented next (Becker 1980), and post-lexical attempts to match prime and target according to prior associations or shared semantic features that are stored within semantic memory (Neely, et al. 1989). The degree to which each of these processes contribute to the semantic priming effect depends on various experimental parameters such as the proportion of related prime-target pairs in a stimulus set (the relatedness proportion, RP) and the time interval between the onset of the prime and the onset of the target (the stimulus onset asynchrony, SOA). It also depends on the nature of the semantic relationship between the prime and the target. For example, when the SOA is relatively long (more than approximately 400msec), and the prime and target are directly related, controlled expectancy and semantic matching processes are the main contributors to the semantic priming effect. However, when the SOA is long and the prime and target are only indirectly related (connected through an unseen mediator word, e.g. “lion” and “stripes”, connected via “tiger”), behavioral semantic priming is usually not seen (Balota and Lorch 1986; Chwilla and Kolk 2002; Chwilla, et al. 2000; Hill, et al. 2002; McNamara and Altarriba 1988; Silva-Pereyra, et al. 1999). This is because an indirectly related target cannot easily be predicted from its prime, and because, in many trials, the semantic relationship between the prime and target is not obvious and the mediator word cannot easily be retrieved from semantic memory and used to bias the lexical decision (discussed by Neely 1991). Interestingly, however, some electrophysiological indirect priming may still be detected, even in the absence of behavioral indirect priming (Chwilla, et al. 2000), perhaps indexing some effects of automatic spreading activation at the neural level.
More recently, several functional neuroimaging studies have described an attenuation of hemodynamic activity across widespread, but variable, regions within temporal and/or the inferior prefrontal cortices in response to directly related, relative to unrelated, word-pairs during semantic priming paradigms (Copland, et al. 2003; Giesbrecht, et al. 2004; Gold, et al. 2006; Matsumoto, et al. 2005; Rissman, et al. 2003; Wheatley, et al. 2005). This attenuation of hemodynamic activity – termed ‘hemodynamic response suppression’ (Henson and Rugg 2003; Henson 2003) – has often been interpreted as reflecting the reduced neurocognitive activity required to process primed targets (Wiggs and Martin 1998) and mirrors the attenuation of the behavioral and electrophysiological responses to primed targets.
There is some variability between fMRI studies in the precise neuroanatomical localization of hemodynamic response suppression observed during semantic priming. Several factors may account for such variability, including the modality of stimulus presentation (visual, e.g. Rossell et al. 2003 versus auditory, e.g. Kotz et al. 2002), the nature of the semantic relationship between prime and target (associative versus categorical, e.g. Kotz, et al. 2002), as well as experimental parameters such as the SOA that bias towards automatic versus controlled processes. For instance, there is some evidence that, at short SOAs, hemodynamic responses suppression within the temporal fusiform cortices (Brodmann area, BA 37) reflects the effects of an automatic spreading activation (Gold, et al. 2006; Wheatley, et al. 2005) while, at longer SOAs, hemodynamic response suppression within the left anterior inferior prefrontal cortex (BA 47) may reflect the facilitory effects of controlled semantic expectations (Gold, et al. 2006). These distinctions, however, are not absolute. For example, Copland et al. (2003) and Wheatley et al. (2005), reported response suppression within the left inferior frontal cortex at short SOAs. And, at long SOAs, several studies have reported modulation within temporal cortices (Gold, et al. 2006; Matsumoto, et al. 2005; Rissman, et al. 2003; Wible, et al. 2006). Moreover, Rossell et al. (2003) failed to show any effect of SOA on suppression within either the inferior prefrontal or fusiform cortices.
Hemodynamic response suppression is not the only pattern of hemodynamic modulation seen in neuroimaging studies of semantic priming using a LD task. In some studies, at long SOAs, it is accompanied by increases in the hemodynamic response to semantically related relative to unrelated word-pairs in other regions (Kotz, et al. 2002; Mummery, et al. 1999; Rossell, et al. 2003; Wible, et al. 2006); indeed, sometimes such increases are the only pattern of modulation observed (Raposo, et al. 2006). These increases in the hemodynamic response to primed relative to unprimed targets are known as hemodynamic response enhancement and are thought to reflect the engagement of neurocognitive processes that occur selectively on primed targets (Henson and Rugg 2003; Henson 2003). There are some consistencies in the localization of such hemodynamic response enhancement reported in semantic priming paradigms: for example, several studies have reported response enhancement within parietal cortices (BAs 40 and 7) (Kotz, et al. 2002; Raposo, et al. 2006; Rossell, et al. 2003; Wible, et al. 2006). This is interesting because such regions constitute part of an attentional circuitry (Behrmann, et al. 2004; Chein, et al. 2003; Cristescu, et al. 2006) that may be specifically engaged as participants attempt to match semantic associations and common semantic features between prime and target after both have been recognized.
In a Relatedness Judgment (RJ) task, participants are explicitly told to search for semantic features or associations that are shared between pairs or groups of words and to use such associations or feature to determine whether or not the words are related to each other. Such a task therefore taps directly into post-lexical semantic matching processes. When participants perform RJ, it takes longer to conclude that a word-pair is unrelated than that it is semantically related (Faust and Lavidor 2003; Zwaan and Yaxley 2003). And, in ERP studies, asking participants to attend to the semantic relationships between prime and target tends to increase the magnitude of the N400 effect produced, both using word stimuli (Holcomb 1988) and picture stimuli (McPherson and Holcomb 1999). More recently, we have also shown that, in comparison with an implicit semantic word-monitoring task, an explicit RJ task also increases the size of the N400 effect produced by indirectly related, relative to unrelated, targets (Kreher, et al. 2006).
In fMRI studies, a RJ task has been used in a variety of different paradigms exploring the functional neuroanatomy underlying semantic processing. In most such studies, however, activity has been collapsed across different types of semantic relationships (Thompson-Schill, et al. 1997; Vandenberghe, et al. 1996), making it difficult to infer which brain regions are recruited in specific association with related versus unrelated word-pairs during explicit semantic processing. Nonetheless, there is some evidence that inferior parietal regions (BA 40) are engaged as participants make explicit similarity judgments about objects with common semantic features (Grossman, et al. 2002; Koenig, et al. 2005). And there is also evidence that, during a RJ task, the left inferior prefrontal cortex (Brodmann Areas 45 and 47) is recruited in association with word-pairs (Fletcher, et al. 2000) and word-triplets (Sabsevitz, et al. 2005) that are more distantly (versus more closely) semantically related.
In summary, there is some evidence from previous neuroimaging studies that hemodynamic response suppression during implicit semantic processing might reflect the effects of a spread of activation under automatic experimental conditions (fusiform suppression), and of predictive strategies under controlled experimental conditions (left inferior frontal suppression), while hemodynamic response enhancement (within left inferior parietal and inferior frontal cortices) might reflect controlled post-lexical matching processes, also occurring under controlled experimental conditions. Such post-lexical matching processes would be maximally engaged as participants carry out explicit semantic relatedness judgments. To date, however, no study has examined the effects of task and semantic relationship on the hemodynamic modulation in the same participants, using the same stimuli and the same experimental parameters.
In the present study, the same participants viewed the same directly related, indirectly related and unrelated word-pair stimuli as they performed both an implicit LD task and an explicit RJ task, using a long SOA that biased towards controlled processing. Both semantic relatedness and task, however, were counterbalanced such that no single participant saw the same word more than once. Based on the findings by Gold, et al. (2006), we predicted that, during the LD task, the directly related (relative to unrelated) word-pairs would be associated with both faster RTs and with hemodynamic response suppression within the left anterior inferior prefrontal cortex (BA 47), reflecting the facilitatory effects of strategic semantic expectancies on processing. We predicted that indirectly related (relative to unrelated) word-pairs would neither be associated with faster RTs nor with left inferior prefrontal cortex response suppression because such expectancy strategies would be ineffective. Based on the findings of Grossman, et al. (2002) and Koenig, et al. (2005), during the explicit RJ task, we predicted that participants’ attention to semantic relationships between the directly and indirectly related word-pairs (relative to the unrelated word-pairs) would be reflected by hemodynamic response enhancement within the left inferior parietal cortex (BA 40). Based on the findings by Fletcher, et al. (2000) and Sabsevitz, et al. (2005), we also predicted that participants’ explicit attempts to retrieve the mediator linking the indirectly related word-pairs would be additionally associated with response enhancement within the left inferior prefrontal cortex.
The stimuli were designed such that they could be counterbalanced both across the two tasks (LD and RJ) and across the three relatedness conditions (directly, indirectly and unrelated). In order to counterbalance in this way, three hundred word triplets were developed such that target words (e.g. “stripes”) were paired with directly related primes (e.g. “tiger”) and indirectly related primes (e.g. “lion”). Word-triplets were taken from those used in previous published studies (Balota and Lorch 1986; McNamara and Altarriba 1988; Weisbrod, et al. 1999) or else were developed for the current study.
In 113 of these triplets, we conducted a free association experiment (described by Kreher et al. 2006, Experiment 3) in which 30 participants who did not participate in the fMRI experiment generated 5 associates to either the primes from the directly related word-pairs, the primes from the indirectly related word-pairs, or to the target words. The directly related targets were almost always generated as associates while the unrelated targets were almost never generated as associates from the primes of the directly related word-pairs. The theoretical mediating words or the primes of the directly related word-pairs were often generated from the primes of the indirectly related word-pairs, while the targets of the indirectly related word-pairs were almost never generated from the primes of the indirectly related word-pairs (for details, see Kreher et al. 2006). In the additional 187 triplets, all the directly related word-pairs, but none of the indirectly and unrelated word-pairs, had an associative strength on the Edinburgh Associative Thesaurus (Coltheart 1981) of greater than zero; and, again, the associative strength of the indirect primes to their theoretical mediating words was much greater than the associative strength between the indirectly related word-pairs.
A second norming study established that, although individuals generally generated the mediator words of the indirectly related word-pairs when given both prime and target, they did not generate the mediator word when they were just given the prime. Finally, fifteen subjects who did not take part in the fMRI experiment conducted a Relatedness Judgement task on the word-pairs in which they were asked to rate how related in meaning they were on a five-point scale using three counterbalanced lists; the directly-related word-pairs (mean = 4.41, SD = 0.56) were rated as being more related in meaning than the indirectly related word-pairs (mean = 3.11, SD = 0.60) [t(299) = 27.207, p < .01], which were, in turn rated as more related in meaning than the unrelated word-pairs (mean =1.45, SD = 0.37) [t(299) = 40.579, p < .01].
These word triplets were then used to counterbalance targets across the six lists in a Latin Square design. Each participant saw one list during the LD task and one list during the RJ task. This ensured that no individual would see the same prime or target more than once (avoiding repetition priming effects), but that, across all participants, exactly the same targets would be seen in all three relatedness conditions in both tasks and that, across all participants, exactly the same primes would be viewed in the directly related and the unrelated conditions. Thus, in each of the six list there were 150 pairs: 50 directly related pairs, 50 indirectly related pairs and 50 unrelated pairs. In addition, the frequency and number of letters of both primes and targets across the six lists (and the three Relatedness conditions) was the same (no main effect of List or no List by Relatedness interaction, p > 0.5). Then, to each list, 50 word-nonword trials were added. All nonword targets were phonologically permissible strings in English and they were all derived from words that were unrelated to their primes. The nonwords were also counterbalanced across the LD and RJ tasks (they were included in the RJ task so that, counterbalanced across all participants, exactly the same stimulus lists could be used in both tasks).
Given that, on the RJ task, participants classified 50% of the indirectly related word-pairs as related and 50% as unrelated (reported in the Results), the RP was approximately 0.5. The nonword ratio (the number of word-nonword-pairs/word-nonword-pairs + unrelated pairs) (Neely, et al. 1989) was 0.4. An example stimulus set is given in Table 1.
Participants were recruited by advertisement. All were right-handed as assessed using the modified Edinburgh Handedness Inventory (Oldfield 1971; White and Ashton 1976). Selection criteria required all participants to have normal or corrected-to-normal vision, to be native speakers of English, and to have learned no other language before the age of five. In addition, volunteers were not taking any medication and were screened to exclude the presence of psychiatric and neurological disorders and to exclude contraindications for MRI. Written consent was obtained from all subjects before participation according to the established guidelines of the Massachusetts General Hospital Institutional Review Board. Two subjects were excluded because of scanning artifacts and one subject was excluded because his behavioral performance was at chance. This left sixteen participants in total (14 males and 2 females; mean age: 42).
During scanning, each participant viewed one list during the LD task and one list during the RJ task (lists were fully counterbalanced across participants as explained above). Each list was divided into three functional runs, each lasting 4 minutes and 10 seconds. The LD task was performed during the first three functional runs, and the RJ task was performed during the second three functional runs. The LD task always took place before the RJ task so that participants were not explicitly alerted to the semantic relationships between the word-pairs that could potentially bias their lexical decisions1.
During the LD task, subjects decided as quickly and as accurately as possible whether the target was a real English word or a nonword. During the RJ task, subjects decided as quickly and as accurately as possible whether the target was related or unrelated in meaning to the prime. Participants were explicitly told that, when they saw target nonwords during the RJ task, they should indicate that these were not related in meaning to the primes. In both tasks, participants indicated their decisions by pressing one or two buttons using the index and middle fingers of their left hand (counterbalanced across subjects). Participants were practiced on the LD task before scanning and on the RJ task inside the scanner after carrying out the LD task. Subjects’ accuracy and reaction times (RTs) on both tasks were recorded.
In both tasks, each trial began with the prime (500msec), a blank screen (300msec), a target (500msec), and then another blank screen (300msec). Thus, the SOA was 800msec. Between word-pairs, a question mark appeared (1100msec) followed by a blank screen (300msec). The four trial types appeared in pseudorandom order, in all runs, interspersed among 100 visual fixation trials (fixate on a “+” for variable durations of 1000msec-8000msec, mean: 3000msec). The random interleaving of these fixation or ‘null-events’ amongst the word-pairs enabled the efficient estimation and deconvolution of the entire hemodynamic response (Burock, et al. 1998).
Subjects underwent two structural scans on a 1.5 Tesla scanner (Siemens Medical Solutions, Iselin NJ), each constituting a 3D MPRAGE sequence (128 sagittal slices, 1.3mm thickness, TR: 7.25msec, TE: 3msec, flip angle: 7°, bandwidth: 195 Hz/pixel, in-plane resolution: 1.3mm × 1mm). Functional imaging took place in a 3.0T head-only Siemens Allegra scanner. Blood oxygen level dependent (BOLD) signal was imaged using a T2*-weighted gradient-echo pulse sequence (TR: 2sec, TE: 25msec, flip angle: 90°) with 33 transverse slices covering the whole brain (125 images per slice, 3mm thickness, 0.9mm between slices). The in-plane resolution was 3.13×3.13mm (64×64 matrix, 200mm FOV). 125 images were acquired during each functional run for a total run time of 4mins 10sec. Head motion was minimized using pillows and a forehead strap. The first four volumes of each functional run were discarded to allow the magnetization to equilibrate.
On the LD task, the frequencies with which nonwords were classified as words (false positive errors) and with which words were classified as nonwords (false negative errors) are reported. On the RJ task, the frequency with which the unrelated words were classified as related (false positive errors) and with which the related words were classified as unrelated (false negative errors) are reported. In addition, the frequency with which the indirectly related words were classified as unrelated are reported. Note that the judgments of the indirectly related word-pairs were subjective – they could be judged as related or unrelated depending on whether, within the time period given, participants were able to retrieve a potential mediator. They therefore cannot be considered correct responses or errors per se.
Given our a priori predictions, we performed planned repeated-measures 2 (Task) × 2 (Relatedness) ANOVAs that contrasted (a) the directly related and the unrelated word-pairs, (b) the indirectly related and the unrelated word-pairs, and (c) the directly related and the indirectly related word-pairs. Planned paired t-tests within the LD or RJ tasks were conducted to examine the source of any interactions between Task and Relatedness. Both subjects analyses (in which RTs were averaged over all items in each relatedness condition) and items analyses (in which RTs were averaged over all subjects in each relatedness condition) were conducted. In both subjects and items analyses, Task and Relatedness were within-subject factors. In all ANOVAs and t-tests, the dependent variable was RTs to the correctly-answered trials: for the LD task, these were the trials on which the targets were correctly classified as words; for the RJ task, these were the trials on which participants classified the directly related and the indirectly related word-pairs as related, and the unrelated word-pairs as unrelated. Because, as discussed above, during the RJ task, the decision as to whether the indirectly related words were related or unrelated was subjective, all analyses that involved the RJ task were repeated (a) including all RTs to indirectly related word-pairs, regardless of how they were classified in the RJ task, and (b) including RTs to indirectly related word-pairs that were judged as unrelated in the RJ task.
Alpha was set to 0.05. All analyses were repeated after logarithmically transforming the data and yielded the same pattern of findings.
In order to increase the signal-to-noise ratio, the two structural scans for each participant were averaged together, after motion correction, to create a single volume2. This resulting high signal:noise volume was then subject to an automated segmentation procedure by which the surface representing the gray/white border was reconstructed and inflated to yield a 2D representation of the cortical surface (Dale, et al. 1999; Dale and Sereno 1993; Fischl, et al. 2001) using FreeSurfer software developed at the Martinos Center, Charlestown, MA (http://surfer.nmr.mgh.harvard.edu/).
Functional images were motion corrected using the AFNI algorithm (Cox 1996; Cox and Jesmanowicz 1999). Images were corrected for temporal drift, normalized and spherically smoothed using a 3D spatial filter (full-width-half-max: 8.7mm). The functional images were then analyzed with a General Linear Model (GLM) using a finite impulse response (FIR) model, using FreeSurfer Functional Analysis Stream (FS-FAST). The FIR model gave estimates of the hemodynamic response every 1sec as stimuli were allowed to onset on half as well as the full 2sec TR. It allowed us to address our hypotheses without assumptions about the shape of the hemodynamic response (Burock, et al. 1998; Burock and Dale 2000; Dale 1999).
The cortical surface of each individual was morphed/registered on to an average spherical surface representation to align sulci and gyri across subjects (Fischl, et al. 1999a; Fischl, et al. 1999b). This structural spherical transform was used to map the GLM parameter estimates and residual error variances of each participant’s functional data to a common spherical coordinate system (Fischl, et al. 1999a; Fischl, et al. 1999b). Each participant’s data was then smoothed on the surface tessellation using an iterative nearest-neighbor averaging procedure, equivalent to applying a two-dimensional Gaussian smoothing kernel with a FWHM of approximately 8.5mm. Because this smoothing procedure was restricted to the cortical surface, averaging data across sulci or outside gray matter was avoided.
BOLD activity to correctly-answered trials was examined in the LD task (i.e. the trials on which the targets were correctly classified as words). In the RJ task, BOLD activity was examined to correctly-answered unrelated and related trials and to the indirectly related trials that were classified as related. However, because relatedness decisions to these indirectly related word-pairs is subjective, we also examined BOLD activity in the RJ task to all indirectly related word-pairs (regardless of how they were classified), as well as to indirectly related word-pairs that that were judged as unrelated. We note any differences in the findings revealed by these different analyses.
Because the LD and RJ tasks may have engaged neural processes at different latencies, we first examined the hemodynamic time courses that were generated during each of these tasks and to each type of word-pair, without any assumption about their overall shapes. These hemodynamic time courses were generated by averaging activity across voxels within temporal and prefrontal regions of interest at each TR (using the FIR model) and across all participants, see Figure 1. The time window that captured the peak of this hemodynamic response across the two tasks and the three relatedness conditions was approximately 3–6 seconds. Therefore, all the statistical maps described below were constructed by summing activity at each voxel across this time-epoch.
We first constructed a statistical map examining the regions activated across both tasks relative to the low-level baseline fixation condition. We also determined whether any of these regions were differentially modulated across the two tasks. We then constructed statistical maps based on planned 2 (Relatedness: directly related versus unrelated, or indirectly related versus unrelated) × 2 (Task: LD versus RJ) repeated measures ANOVAs to show the main effects of Relatedness as well as Task by Relatedness interactions. In these analyses, only ‘highest order’ effects are shown/reported. In other words, clusters that we report as showing main effects for a particular factor are those that failed to show Task by Relatedness interactions. In order to determine the sources of any significant Task by Relatedness interactions as well as to examine hemodynamic modulation by semantic relationship within the LD and RJ tasks within regions that did not show significant main effects or interactions in the overall ANOVA maps, we also constructed statistical maps comparing the directly related and unrelated word-pairs and comparing the indirectly related and unrelated word-pairs for the LD and RJ tasks separately. Finally, we constructed statistical maps that directly contrasted the directly and indirectly related word-pairs for each of the LD and RJ tasks. This enabled us to determine the specificity of any hemodynamic modulation to directly versus indirectly related word-pairs.
Correction for multiple comparisons depended on whether voxels fell within or outside a priori regions of interest (see Figure 1). Within regions of interest (Figure 1), p values for sets of contiguous voxels (clusters) were computed using a permutation (Nichols and Holmes 2002) with 10000 iterations; a cluster was only considered significant if, on this permutation, its significance was less than p = 0.05. These clusters are indicated with a * in Tables 6 and and7.7. Outside regions of interest, we also report clusters that covered at least 300mm2, with a corrected threshold for rejection of the null hypothesis of p < 0.05, identified on the basis of a Monte Carlo simulation across the whole cortical surface (Doherty, et al. 2004). These clusters are indicated with a # in Tables 6 and and77.
There were no differences in errors (falsely classifying a nonword as a word or erroneously classifying words as nonwords) in the LD task across the four conditions, F (3, 42) = 1.50, p = 0.24. A′ prime scores in all participants in all conditions were more than 0.8, suggesting that there were no response biases. In the RJ task, there was no significant difference in errors to the related word-pairs (falsely classifying them as unrelated) and the unrelated word-pairs (falsely classifying them as related), t(15) = 1.221, p = 0.241. The range of errors in the LD task was between 0–8% and in the RJ task was between 0–20%, with the exception of one participant who had, on average 26% errors in the LD task and 33% errors in the RJ task. Exclusion of this participant made no difference to the pattern of behavioral findings reported in Table 4. The judgment of the indirectly related word-pairs was subjective and depended on whether participants were able to identify the mediating word in the time-interval given: on average, 50% of the indirectly related word-pairs were classified as unrelated (range: 28–94%).
Comparison between the unrelated and the directly related word-pairs revealed a main effect of Relatedness and a Task by Relatedness interaction (Table 4). Follow ups showed that the direct priming effect was greater in the RJ task (significant on both items and subjects analyses, ps < 0.0001) than in the LD task (significant on the items analysis, p < 0.05 but not on the subjects analysis3).
Comparison between the unrelated and the indirectly related word-pairs (in the RJ task, those indirectly-related word-pairs that were classified as related) revealed a Task by Relatedness interaction that approached significance on the subjects analysis and that reached significance on the items analysis. Follow-ups failed to show priming effects on the LD task but showed reverse priming effects on the RJ task with longer RTs to the indirectly related than to the unrelated word-pairs that reached significance on the items analysis4.
A direct comparison between RTs to the directly and indirectly related word-pairs also revealed significant Task by Relatedness interactions, with follow-ups confirming longer RTs to the indirectly related word-pairs than to the related word-pairs on the RJ task (ps < 0.001) but no significant differences on the LD task.
Finally, all three 2 × 2 ANOVAs all showed main effects of Task due to longer RTs in the RJ task than in the LD task.
As expected, a large network was activated (Table 5A) and another network was deactivated (Table 5B) in comparing all word-pairs with the fixation condition (Figure 2). Additionally, a few of these regions were modulated by task (Table 6). Of most interest, however, were the comparisons between the three relatedness conditions.
In contrasting the directly related with the unrelated word-pairs, a cluster within the left inferior temporal/fusiform gyrus and a cluster at the right occipito-parietal junction showed main effects due to response suppression across both the LD and RJ tasks (Table 7A, Figure 3A). These clusters failed to show Task by Relatedness interactions, suggesting that they were modulated to the same degree across both tasks. In addition, a left anterior inferior prefrontal and a right lateral orbitofrontal cluster showed hemodynamic response suppression during the LD task (Table 7A, Figure 3A). These two clusters did not show main effects of Relatedness across both tasks or Task by Relatedness interactions.
In contrasting the indirectly related with the unrelated word-pairs, there was fairly widespread hemodynamic response suppression within inferior temporal, occipital, parietal and cingulate cortices during the LD task (Table 8A, Figures 3B). Some of these regions also showed main effects of Relatedness across the LD and RJ tasks, but none showed main effects on the RJ task alone and none showed Task by Relatedness interactions.
A direct comparison between the indirectly and the directly related word-pairs indicated that the left lateral anterior inferior temporal cortex and left anterior cingulate clusters showed significantly more suppression to the indirectly related than to the directly related word-pairs (indicated with ** in Table 8A).
During the RJ task, other regions showed response enhancement (more activity to the directly or indirectly related word-pairs than to the unrelated word-pairs). In comparing both the directly and the indirectly related word-pairs with the unrelated word-pairs, a cluster within the left inferior parietal lobule showed response enhancement during the RJ task, but no significant modulation during the LD task. This difference in modulation across the two tasks was reflected by a Task by Relatedness interaction (Tables 7B and and8B,8B, Figure 4). A direct comparison between the related and the unrelated word-pairs did not reveal any difference in modulation within this cluster.
In addition, comparing the indirectly related word-pairs with the unrelated word-pairs during the RJ task, response enhancement was also seen within the left inferior frontal cortex (Table 8B, Figure 4B). A direct comparison between the indirectly and directly related word-pairs during this task confirmed that this region showed more activity to the indirectly related word-pairs than to the directly related word-pairs (indicated with ** in Table 8B).
This study contrasted the effects of an implicit and an explicit task on the modulation of the hemodynamic response to the same directly related and indirectly related word-pairs, relative to unrelated word-pairs. In the implicit LD task, participants simply decided whether a target word was a word or a nonword. In the explicit RJ task, participants determined whether or not the prime and target were related in meaning. The results were striking. The task affected both the polarity of hemodynamic modulation as well as the neuroanatomical regions that were modulated. The semantic relationship between the word pairs (direct or indirect) affected the neuroanatomical regions that were modulated. The LD task led to hemodynamic response suppression within bilateral anterior inferior prefrontal cortices (BA 47 on the left; BA 10/11 on the right) and within the left inferior temporal/fusiform cortex (BA 37) to directly related (relative to unrelated) word-pairs, and to hemodynamic response suppression within more widespread temporal and occipital regions to indirectly related (relative to unrelated) word-pairs. The RJ task, on the other hand, was mainly associated with hemodynamic response enhancement. This response enhancement was observed within the left inferior parietal lobule (BA 40) to both directly and indirectly related word-pairs as well as within the left inferior prefrontal cortex (BAs 47 and 45) to indirectly related word-pairs, each relative to unrelated word pairs. Below we consider these patterns of hemodynamic response suppression and enhancement in relation to participants’ behavioral responses, the potential cognitive processes engaged, and previous neuroimaging studies of semantic processing.
We predicted that, during LD task, the left anterior inferior prefrontal cortex (BA 47) – a region known to mediate controlled semantic retrieval processes (Wagner, et al. 2001) – would show response suppression to the directly related relative to unrelated word-pairs. This prediction was confirmed. Response suppression within the left inferior prefrontal cortex is consistent with several previous neuroimaging studies of direct semantic priming using a LD task (Copland, et al. 2003; Giesbrecht, et al. 2004; Gold, et al. 2006; Matsumoto, et al. 2005; Wheatley, et al. 2005; Wible, et al. 2006). Consistent with the interpretation of Gold et al. (2006), we suggest that its response suppression reflected the relative ease of accessing target words that had been predicted from their directly related primes through controlled semantic expectancy strategies (Neely 1991). In the current study, the additional response suppression within the right anterior orbitofrontal cortex (BA 10/11) to related (relative to unrelated) is less consistent with previous fMRI studies of semantic priming, but may reflect the more general involvement of right inferior prefrontal regions in inhibitory processes (Aron, et al. 2004), possibly in inhibiting predictions that did not match unrelated targets.
A failure of inferior prefrontal suppression and the absence of a significant behavioral priming effect to the indirectly related word-pairs, relative to unrelated word-pairs, was also predicted. The absence of indirect priming under controlled experimental conditions with long SOAs (Balota and Lorch 1986; Chwilla and Kolk 2002; Chwilla, et al. 2000; Hill, et al. 2002; McNamara and Altarriba 1988; Silva-Pereyra, et al. 1999) has been explained by positing that any expectancy strategies in which participants engage are just as ineffective in predicting indirectly related targets as in predicting unrelated targets (Neely 1991). If, as discussed above, hemodynamic response suppression within the inferior and ventral prefrontal cortices reflects the reduced retrieval effort that results from such predictions, this would explain why it was not suppressed to the indirectly related word-pairs: it was engaged to the same degree as to the unrelated word-pairs.
Interestingly, despite the long SOA, temporal fusiform cortices (BA 37) that previous studies have implicated in the storage of lexico-semantic representations (Nobre and McCarthy 1995; Price 2000; Van Petten and Luka 2006) and their automatic access through processes such as spreading activation (Gold, et al. 2006; Wheatley, et al. 2005), also showed hemodynamic response suppression during the LD task in response to the directly related, relative to the unrelated, word-pairs. Response suppression within the temporal fusiform cortex has been previously described at long SOAs (Gold, et al. 2006; Matsumoto, et al. 2005; Rissman, et al. 2003; Wible, et al. 2006; Rossell et al. 2003). Although, under these circumstances, any spreading activation is unlikely to contribute to behavioral priming, it is still possible that BOLD suppression within this region was been sensitive to the effects of spreading activation at the neural level.
In comparing the indirectly related and unrelated word-pairs, hemodynamic suppression was not confined to the temporal fusiform cortex, but was also observed within other temporal-occipital regions including the right medial temporal cortex and the left lateral anterior temporal cortex, and within bilateral extrastriate cortices. Although there was no indirect behavioral priming during the LD task, it is still possible that this hemodynamic suppression to the indirectly related word-pairs still reflected a spread of activation that was picked up at a neural level. This would be consistent with a previous report of some neurophysiological priming in the absence of behavioral indirect priming (Chwilla, et al. 2000). Indeed, the relatively widespread suppression may have reflected the longer time such activation had to build up, spread and hemodynamically prime lexico-semantic representations stored within these cortices before participants made their lexical decisions.
During the RJ ask, hemodynamic response enhancement was observed within the left inferior parietal lobule (BA 40) in response to both the directly and the indirectly related, relative to the unrelated, word-pairs. We suggest that the recruitment of these regions reflected participants’ attention to semantic associations or common semantic features between prime and target, as they attempted to find semantic matches between them5. This interpretation accords with the findings of previous studies that have reported parietal activation in association with the acquisition and application of semantic categorical rules to classify novel objects with common semantic features, as well as with its activation as participants make similarity judgments about objects with common semantic features (Grossman, et al. 2002; Koenig, et al. 2005). It is also consistent with views that the left inferior parietal cortex, plays a role in attentional focus and shifting (Behrmann, et al. 2004), aspects of working memory maintenance (Ravizza, et al. 2004), and the integration of semantic features across words (Grossman, et al. 2002; Koenig, et al. 2005). Indeed, there is evidence that some of these functions may be related. For example, in addition to its known role in spatial attention (Corbetta and Shulman 2002), the left parietal cortex is recruited when participants are specifically cued to attend to semantic attributes of a word target (Cristescu, et al. 2006).
In addition to hemodynamic response enhancement within the inferior parietal cortex, the indirectly related word-pairs also led to hemodynamic response enhancement within the left inferior frontal cortex (BA 47 and 45). This is consistent with previous studies that have demonstrated response enhancement within the left inferior prefrontal cortex in association with semantic relatedness decisions on word-pairs (Fletcher, et al. 2000) and word-triplets (Sabsevitz, et al. 2005) that were more distantly (versus more closely) semantically related. Unlike the contrast between the directly related and unrelated word-pairs, RTs to the indirectly related word-pairs were longer than to the unrelated word-pairs. We suggest that both these longer RTs and the recruitment of the left inferior prefrontal cortex to the indirectly related word-pairs reflected participants’ attempts to retrieve the specific mediators linking indirectly related primes and targets in order to make their relatedness judgments.
Notably no response enhancement, either to the directly or to the indirectly related (relative to the unrelated) word-pairs, was observed during the LD task. This contrasts with some other fMRI semantic priming studies that have used a LD task and that have reported some hemodynamic response enhancement in addition to, or even instead of, response suppression (Kotz, et al. 2002; Mummery, et al. 1999; Raposo, et al. 2006; Rossell, et al. 2003; Wible, et al. 2006). In the current study, the absence of hemodynamic response enhancement during the LD task does not imply that post-lexical semantic matching processes were not contributing to priming at all. However, it is possible that these matching processes were not operating to the same degree as in some of these previous fMRI studies. First, in the present study, the RP was relatively low (approximately 0.5 given that 50% of the word-pairs were classified as related) and this may have reduced the efficacy of any semantic matching processes in speeding up lexical decisions to primed targets. Second, although the SOA in the current study (800msec) was long enough to allow some semantic matching, it was not as long as in a study by Raposo et al. (2006) that reported only response enhancement to related (relative to unrelated) word-pairs during a LD task and in which the 2500msec between prime and target is likely to have encouraged participants to attend to any semantic relationships between them. Future studies will determine whether experimental parameters such as RP and SOA can predict the degree of hemodynamic response enhancement to related relative to unrelated word-pairs during an implicit LD task.
In sum, the current study demonstrates that, as the same participants viewed the same directly related, indirectly related and unrelated word-pairs, the task they performed influenced both the polarity and neuroanatomical localization of hemodynamic modulation. We have suggested that, during the LD task at a long SOA, inferior prefrontal response suppression reflected participants’ successful predictions of directly related targets and that this was the primary determinant of the behavioral priming effect. We have also suggested that the temporal (and occipital) hemodynamic response suppression reflected some effect of spreading activation across stored neural word representations, even though, at this SOA, such automatic activation did not contribute significantly to behavioral priming. Finally, we have suggested that, during the RJ task, the left parietal response enhancement reflected participants’ attention to semantic relationships as they attempt to semantically match primes and targets, and that the additional left inferior prefrontal response enhancement to the indirectly related word-pairs reflected participants’ attempts to retrieve the specific words that mediated between indirectly related primes and targets.
Although the explanations of these patterns of response enhancement and suppression are still relatively hypothetical, they lead to specific predictions about the time course of activation within these regions during semantic priming under controlled experimental conditions: they predict that response suppression due to pre-lexical automatic spreading activation and controlled expectancy generation, will occur before response enhancement due to post-lexical semantic matching processes. Such hypotheses cannot be tested directly using fMRI that has an inherently poor temporal resolution. However, it may be possible to combine its excellent spatial resolution with techniques such as ERPs and magneto-encephalography (MEG) that do have the temporal resolution to examine the precise time courses of these neurocognitive processes. Encouragingly, there is already some convergence of findings across studies that have used fMRI and ERP/MEG techniques. For example, intracranial ERP studies have implicated the anterior fusiform cortex (Nobre and McCarthy 1995; Halgren, et al. 1994b) as well as the left inferior prefrontal cortex (Halgren, et al. 1994a) – both regions that were modulated in the current study – as sources of the N400 evoked in single word paradigms. And MEG studies have also demonstrated modulation within temporal and inferior prefrontal cortices within the N400 time window during word repetition priming (Marinkovic, et al. 2003) and sentence anomaly (D’Arcy, et al. 2004; Halgren, et al. 2002; Helenius, et al. 1998) paradigms. Future studies combining the spatial resolution of fMRI with the temporal resolution of ERP and MEG techniques (Dale, et al. 2000) will be able to test the model outlined in the present study more directly.
This research was supported by NIMH (R01 MH071635, K01 MH65356) and the Institute for Mental Illness and Neuroscience Discovery (MIND). Gina R Kuperberg was also supported by NARSAD (with the Sidney Baer Trust) and by a Claflin Distinguished Scholars Award from Mass. General Hospital. We thank Kristin Girasa for her help in developing the stimuli, Thilo Deckersbach and Daphne Holt for their help with scanning, and Sarah Groff for her help with data analysis.
1It is theoretically possible that a systematic difference in noise between the first three and the last three runs could have confounded our assessment of hemodynamic modulation across the two tasks. We therefore computed the difference in the residual variance at each voxel (a direct measure of both scanner and physiological noise) between the LD task (the first three runs) and the RJ task (the last three runs) in each participant, averaged this difference across all participants in standardized space and then used t-tests to determine whether these differences were significant at any voxel. This analysis failed to reveal any significant differences in noise at any voxel across the cortex at p < 0.01, uncorrected for multiple comparisons.
2We now use a higher sensitivity coil to acquire these high-resolution structural scans, and signal averaging is therefore no longer necessary to achieve the required signal:noise for the automated reconstruction procedures.
3Behavioral data collected from a larger sample of controls (n=36) outside the scanner using exactly the same paradigm, however, did reveal a significant behavioral direct priming effect on the subjects analysis t(35) = 2.8, p < 0.009.
4When ANOVAs were repeated including RTs to indirectly related word-pairs that were judged as unrelated in the RJ task or using all RTs to indirectly related word-pairs regardless of how they were classified in the RJ task, both Task by Relatedness interactions and reverse priming effects in the RJ task reached significance on both subjects and items analyses.
5Our piloting studies indicated that, even though participants failed to generate indirectly related targets to primes on a word association task, they recognized some semantic relationship between them and tended to rate the indirectly related word-pairs as more related than the unrelated word-pairs.