## Abstract

Meaning retrieval of a word can proceed fast and effortlessly or can be characterized by a controlled search for candidate lexical items and a subsequent selection process. In the current study, we facilitated meaning retrieval by increasing the number of words that were related to the final target word in a triplet (e.g., lion–stripes–tiger). To induce higher search and selection demands, we presented ambiguous words as targets (i.e., homonyms like ball) in half of the trials. Hereby, the dominant (game), low-frequent (dance), or both meanings of the homonym were primed. Participants performed a relatedness judgment during functional magnetic resonance imaging. Activation in a bilateral network (angular gyrus, rostromedial prefrontal cortex) increased linearly with multiple related primes, whereas the posterior left inferior prefrontal cortex (pLIPC) showed the reverse activation pattern for unambiguous trials. When homonyms served as targets, pLIPC responded strongest when both meanings or low-frequent concepts were addressed. Additional anterior left inferior prefrontal cortex activation was observed for the latter trials only. The data support an interaction between 2 distinct cerebral networks that can be linked to automatic bottom-up support and top-down control during meaning retrieval. They further imply a functional specialization of the LIPC along an anterior–posterior dimension.

## Introduction

The human mind resembles a complex information processing system based on multiple functionally distinct neural components that interact with each other in order to achieve a desired behavior, for example, speech. With respect to language, lesion studies provided early evidence for a spatiofunctional dichotomy between representational and executive components of language processing in the human brain. Damage to the left inferior prefrontal cortex (LIPC) resulted in substantial speech impairments despite intact semantic knowledge and indicated, together with recent neuroimaging data, the relevance of the LIPC for high-level executive processes like semantic search, strategic access, inhibition, and selection (for a review, see Thompson-Schill et al. 1997, 1998, 1999, 2003; Wagner et al. 2001; Bookheimer 2002; Badre et al. 2005). In contrast, activation associated with semantic knowledge representation appears to include structures in the temporal and/or parietal lobes (e.g., Gold and Buckner 2002; Gold et al. 2006; Noppeney et al. 2007), although current feature-based models propose a more distributed conceptual network on the basis of a word's sensory–motor properties (for reviews, see Barsalou et al. 2003; Martin 2007). Despite different opinions on how and where knowledge is stored, substantial support is provided that access to semantic knowledge is mediated by a top-down influence of the LIPC onto conceptual knowledge areas (Thompson-Schill et al. 1999; Gold and Buckner 2002; Badre et al. 2005; Gennari et al. 2007; Martin 2007; for alternative views, see Noppeney et al. 2004; Jefferies and Lambon Ralph 2006). When contextual information is weak or the number of competing semantic concepts is high, retrieval based on stimulus-driven, bottom-up mechanisms is insufficient, and a more controlled top-down approach is required (Thompson-Schill et al. 1997; Wagner et al. 2001; Badre et al. 2005). In previous studies, top-down/bottom-up processes were manipulated by varying retrieval, selection, or competition demands while participants made decisions on the relation between pairs of words. Little interest has so far been paid to the amount of prior relevant, contextual information onto meaning retrieval of single words and, more importantly, in how far context can modulate bottom-up or top-down related processes.

In everyday communication, words rarely appear in isolation but are embedded into context. It is generally accepted that the linguistic environment of a target word, whether this involves the prior presentation of single words, syntactic discourse, or prosodic information, biases processing of subsequent material. For example, semantic priming (i.e., faster reaction times [RTs] to a target word when it is preceded by a semantically related relative to an unrelated word) implies the relevance of prior contextual information for target processing (for a review, see Neely 1991). Behavioral investigations with multiple primes have also shown that the amount of context provided previously has an impact on how easy the conceptual representation of a word can be recovered (e.g., Beeman et al. 1994; Faust and Lavidor 2003). Meaning retrieval latencies gradually increased from contexts in which 2 semantically related words (lion–stripe–tiger) preceded the target word, to single related (lion–bread–tiger), to unrelated contexts (rest–bread–tiger) (Balota and Paul 1996; Chwilla et al. 2003; Milberg et al. 2003; Kandhadai and Federmeier 2007). Models of semantic memory postulate that previous information associated with the target enhances its preactivation level and prepares it for subsequent manipulation (Collins and Loftus 1975). A respective decrease of blood oxygen level–dependent (BOLD) activity for related relative to unrelated word pairs (semantic suppression) has been observed in brain regions associated with controlled or effortful retrieval (e.g., LIPC) (Kotz et al. 2002; Copland et al. 2003; Gold et al. 2006; Kuperberg et al. 2008; Ruff et al. 2008). Semantic priming, however, is not solely reflected in reduced BOLD activity. For example, Raposo et al. (2006) did not report semantic suppression at all in their semantic priming experiment. The opposite response, BOLD enhancement, was evident in posterior semantic regions including the inferior parietal cortex (IPC) (Rossell et al. 2003; Raposo et al. 2006; Kuperberg et al. 2008; Ruff et al. 2008) and posterior middle/superior temporal gyrus (Kotz et al. 2002; Raposo et al. 2006). Increased activity has hereby been linked to successful postlexical integration particularly in the IPC (Rossell et al. 2003; Kuperberg et al. 2008). Thus, varying the number of semantically related words preceding the target is likely to address parietal and/or temporal brain regions and seems to be a candidate method for evoking bottom-up processes during meaning retrieval.

Contextual information, despite being associated with the target, can also increase retrieval demands and provoke a higher level of top-down control. Homonyms, for example, can require additional cognitive resources related to the selection of the appropriate meaning among its competitors and the inhibition of irrelevant concepts. Ambiguity studies have shown that RT or eye fixation duration at a homonym (bank) depends on whether the high-frequent, dominant (financial institution) or alternative, subordinate (river side) meaning is addressed compared with an unambiguous control word (“subordinate bias effect,” Rayner et al. 1994). Although the appropriate meaning is successfully retrieved at some stage, the initial influence of contextual and lexicosemantic factors (i.e., word frequency) is still controversially discussed. Unlike traditional models of exhaustive (e.g., Onifer and Swinney 1981) or selective, context-appropriate, access (e.g., Simpson 1981; Martin et al. 1999), hybrid models favor an early interaction of contextual and lexicosemantic factors (e.g., “graded salience hypothesis,” Giora 1999; “reordered access model,” Duffy et al. 1988; Rayner and Frazier 1989; Peleg et al. 2001; Sereno et al. 2003). Both meanings of a homonym compete for selection in subordinate-biasing sentences because context is able to facilitate the appropriate weaker concept but, at the same time, unable to override the dominant meaning completely even at 250-ms postonset of the homonym (Peleg and Eviatar 2008). The additional processing costs required for resolving lexical ambiguities have again been linked to increased LIPC activation (Rodd et al. 2005; Davis et al. 2007; Gennari et al. 2007; Mason and Just 2007; Zempleni et al. 2007). Homonyms, therefore, seem to be a valuable tool to manipulate selection and competition demands during multiple priming.

The aim of the current study was to explore the neural basis of bottom-up and top-down related processes during meaning retrieval using a multiple prime paradigm. Because the 2 mechanisms seem to be closely interrelated, that is, the more bottom-up the less top-down regulation, increases and decreases of BOLD signals were identified using a regression analysis with the number of related primes as covariates. We hypothesize that brain areas showing an upregulation of BOLD activity as the number of related primes increase support bottom-up retrieval and are associated with a successful and effortless meaning integration. In contrast, brain areas showing BOLD enhancements as the number of related primes decrease (i.e., LIPC) mediate a more controlled retrieval, either due to less contextual facilitation (for unambiguous targets) or an increase in competition and selection demands (for homonyms). Direct comparison between unambiguous and ambiguous trials across conditions should further specify components in the top-down network that are specifically related to the additional processing costs during ambiguity resolution, such as high selection and inhibition demands. These should be most evident for subordinate-biasing contexts.

## Materials and Methods

### Participants

In all, 18 male subjects took part in the functional magnetic resonance imaging (fMRI) study. Due to motion-related image artifacts, data from 3 subjects were discarded. Image data from the remaining 15 subjects were analyzed. All participants (mean [M] age = 28.93 years, standard deviation [SD] = 7.11; M years of education = 12.60 years, SD = 1.92) were native German speakers, right handed according to the Edinburgh Inventory of Handedness (Annett 1970) and showed average or above average estimated verbal IQ as assessed by the German MWT-B multiple choice vocabulary test (Lehrl et al. 1995) (M estimated verbal IQ = 111.93, SD = 14.55). Working memory span was measured with the digit span task as part of the Wechsler Adult Intelligence Scale (Tewes 1991) (accumulated score of the forward and backward version: M = 16.13, SD = 4.76) and vocabulary size with lexical verbal fluency (words generated in response to 4 different letters; accumulated score: M = 49.71, SD = 13.46). Subjects were excluded on grounds of recent substance use and neurological or known medical disorders that affect cerebral metabolism or general MRI incompatibility. All participants had normal or corrected-to-normal vision, gave informed consent, and were paid 15 Euros for participation in the study. The study was approved by the local ethics committee.

During the multiple priming tasks, participants saw word triplets (wine–seed–grape) and had to indicate whether the last word (target) was related to at least one of its preceding words (primes) by pressing a button with their left index (“yes”) or middle (“no”) finger. Targets were either ambiguous (bank) or unambiguous (grape). Each word of the triplet was presented individually with an interstimulus interval (ISI) of 0 ms. Primes were shown for 200 ms each and targets for 1000 ms in the centre of the visual field. A stimulus-onset asynchrony of 200 ms was chosen to identify retrieval-related processes without the influence of deliberate strategies or expectancies, to be able to obtain disambiguating competition and selection processes (Duffy et al. 1988; Simpson 1994), and to avoid priming effects for indirectly related primes in the double related trials (Spitzer et al. 1993). The relatedness judgment task ensured conceptual/semantic rather than lexical priming because participants were required to select one meaning of the homonym in order to make their relatedness decision (Balota and Paul 1996; Chwilla and Kolk 2003).

Prime–target relations were systematically manipulated, that is, the target was related to none of the primes in unrelated conditions (UUs), to either the first (RU) or the second prime (UR) in single related conditions and to both primes in double related conditions (RR). None of the primes were directly related to each other in order to determine the independent effect of each prime onto target processing. Besides varying the number of related primes, target processing was further manipulated by dominance compatibility of the primes. Related primes in single related trials either addressed the dominant (Rd) or subordinate meaning (Rs) or both concepts of the homonym (RRa) in the double related condition. The respective unambiguous conditions could not be varied in this way, but unambiguous word triplets were constructed (R1, R2, RR) analogous to the ambiguous conditions. Hereby each prime, R1 and R2, was equally strongly related to the target concept. Finally, to avoid position effects, each prime appeared once in first and once in second position before the target. Table 1 provides examples of each condition.

Table 1

Example of word triplets used in the multiple prime paradigm for double related (RR), single related (R), and unrelated (UU) trials (translated into English)

 Unambiguous Ambiguous Type Prime 1 Prime 2 Target Type Prime 1 Prime 2 Target RR R1R2 Lion Stripe Tiger RRa RdRs Game Dance Ball R2R1 Stripe Lion Tiger RsRd Dance Game Ball R1 R1U1 Lion Bread Tiger Rd RdU1 Game Pillow Ball U1R1 Bread Lion Tiger U1Rd Pillow Game Ball R2 R2U2 Stripe Rest Tiger Rs RsU2 Dance Clock Ball U2R2 Rest Stripe Tiger U2Rs Clock Dance Ball UU U1U2 Bread Rest Tiger UUa U1U2a Pillow Clock Ball U2U1 Rest Bread Tiger U2U1a Clock Pillow Ball
 Unambiguous Ambiguous Type Prime 1 Prime 2 Target Type Prime 1 Prime 2 Target RR R1R2 Lion Stripe Tiger RRa RdRs Game Dance Ball R2R1 Stripe Lion Tiger RsRd Dance Game Ball R1 R1U1 Lion Bread Tiger Rd RdU1 Game Pillow Ball U1R1 Bread Lion Tiger U1Rd Pillow Game Ball R2 R2U2 Stripe Rest Tiger Rs RsU2 Dance Clock Ball U2R2 Rest Stripe Tiger U2Rs Clock Dance Ball UU U1U2 Bread Rest Tiger UUa U1U2a Pillow Clock Ball U2U1 Rest Bread Tiger U2U1a Clock Pillow Ball

Note: R, related; U, unrelated; a, ambiguous; d, prime addressing the dominant meaning of the homonym; s, prime addressing the subordinate meaning of the homonym.

### Stimuli

The unambiguous stimuli were based on the indirectly related word pairs from the study of Spitzer et al. (1993). They consisted of 25 indirectly related prime–target pairs (wine–seed) that were mediated by a common associate (grape). In the present experiment, the mediator served as the target. The ambiguous stimuli were chosen from the association norms for German homonyms (Moritz et al. 2001). All homonyms were primary homonyms with 2 distinct meanings and matched for polarity, that is, there was a clear dominance for one of the meanings with at least 20% difference between the dominant and the subordinate meaning, whereas at least 5% of the spontaneous associations produced for each homonym were related to the subordinate meaning. No differences emerged between ambiguous targets, unrelated or related primes, and their respective unambiguous conditions according to the following criteria: 1) length in syllables (≤4) and letters (≤9), frequency as assessed by the CELEX database (Baayen et al. 1993) and relatedness score (unrelated < 3; related > 5). The latter score was the result of a rating task during which participants (n = 12), not taking part in the fMRI experiment, had to judge in how far 2 words were related to each other on a scale from 1 (=not related) to 7 (=related). A high rating score was required for all related primes and their corresponding targets (rt) but undesirable between 2 related primes (rr) and all unrelated word pairs (ru, ut, uu). The results of the 2 × 5 repeated-measures analysis of variance (ANOVA) with ambiguity (ambiguous, unambiguous) and relation (rr, rt, ru, ut, uu) as within-subject factors indicated that the stimuli chosen for the experiment fulfilled this requirement. The only significant effect was observed for relation (F4,100 = 591.374, P < 0.001). Lowest rating scores were obtained when prime–target pairs were unrelated (M ru = 1.395, SD = 0.467; M ut = 1.308, SD = 0.366; M uu = 1.350, SD = 0.443), highest scores for related pairs (M rt = 6.420, SD = 0.602), and intermediate scores for indirectly related word pairs (M rr = 2.586, SD = 1.333). All between-group differences were significant (P < 0.001). Word pairs in the rr condition were considered to be unrelated although their relatedness values were slightly higher than the unrelated pairs due to the fact that they converge onto a common, distant association.

The 400 word triplets (16 conditions × 25 trials) were split up into 4 lists of 100 trials each. Each list was controlled for the number of target repetitions (each target appeared twice in each list), the number of trials per condition (6 trials), and the way they appeared in succession, that is, once in isolation, once in a set of 2, and once in a set of 3 successive presentations. The remaining 4 slots of each list were filled individually. To avoid sequence effects, trials were pseudorandomized in each of the 4 lists, and the order in which the lists were presented was also randomized and counterbalanced across subjects. Each participant saw all 400 word triplets once.

### fMRI Procedure

A rapid event-related design was used to present the 16 conditions. Small blocks of each condition were created, consisting of 2 or 3 consecutive trials of the same condition. These blocks were randomly interspersed with isolated trials of a different condition. The intertrial interval (ITI) was predominantly short (1–2 s) between uniform sequences and longer (2.5–5 s) between different conditions. This manipulation was used to enhance BOLD signal strength during event-related designs (Amaro and Barker 2006; see also Saβ et al. 2008). Lasting about 32 min in total, participants performed the multiple priming tasks in 4 separate scanning sessions.

At the beginning of each trial, a fixation cross appeared for 500 ms in the centre of the screen. Visually presented primes and targets followed and were shown individually with an ISI of 0 ms. Each prime was presented for 200 ms and the target for 1000 ms. The appearance of the hash symbol indicated the end of the trial and was shown for the entire ITI duration.

Upon appearance of the target, the participant had to decide as fast as possible whether he/she detected a relation between the target and at least one of the preceding primes by pressing a button.

Presentation of stimuli was controlled by a computer using the Presentation 10.1 software package (Neurobehavioral Systems, Inc., Albany, CA, http://www.neurobs.com/). MRI-compatible goggles (VisuaStim XGA; Resonance Technology, Inc., http://www.mrivideo.com/) were used for stimuli presentation. Errors, misses, and outliers were excluded from later event-related analysis.

All subjects performed 10 practice trials and were given feedback before participating in the fMRI experiment. All items used for training differed from the experimental stimuli of the fMRI version.

### MRI Acquisition

Imaging was performed at 1.5 T (Gyroscan Intera; Philips Medical Systems, Best, The Netherlands) using standard gradients and a circularly polarized phase array head coil. For each subject, we acquired 4 series of functional volumes of $T2*$-weighted axial echo-planar imaging (EPI) scans parallel to the anterior and posterior commissure (AC/PC) line with the following parameters: number of slices, 31; slice thickness), 3.5 mm; interslice gap, 0.35 mm; matrix size, 64 × 64; field of view, 240 × 240 mm; time echo, 30 ms; time repetition, 2.8 s. Four runs were acquired with 167 functional volumes each.

### fMRI Data Analysis

MR images were analyzed using Statistical Parametric Mapping software (SPM5) implemented in MATLAB 7.0 (Mathworks, Inc., Sherborn, MA). All images were realigned to the first image to correct for head movement. Unwarping was used to correct for the interaction of susceptibility artifacts and head movement. After realignment and unwarping, the signal measured in each slice was shifted relative to the acquisition time of the middle slice using a sinc interpolation in time to correct for their different acquisition times. Volumes were then normalized into standard stereotaxic anatomical Montreal Neurological Institute (MNI) space by using the transformation matrix calculated from the first EPI scan of each subject and the EPI template. Afterward, the normalized data with a resliced voxel size of 4 × 4 × 4 mm was smoothed with a 10-mm full width at half maximum isotropic Gaussian kernel to accommodate intersubject variation in brain anatomy. The time series data was high-pass filtered with a high-pass cutoff of 1/128 Hz. The autocorrelation of the data was estimated and corrected for.

The expected hemodynamic response at stimulus onset for each event type was modeled by 2 response functions, a canonical hemodynamic response function (HRF) (Friston et al. 1998) and its temporal derivative. The temporal derivative was included in the model to account for the residual variance resulting from small temporal differences in the onset of the hemodynamic response, which is not explained by the canonical HRF alone. The functions were convolved with the event train of stimulus onsets to create covariates in a general linear model. The volume of interest was restricted to gray matter voxels by use of an inclusive mask created from the segmentation of the standard brain template. Subsequently, parameter estimates of the HRF regressor for each of the different conditions were calculated from the least–mean squares fit of the model to the time series. Parameter estimates for the temporal derivative were not further considered in any contrast.

Using SPM5, second-level statistics were calculated in several steps. First, to identify the effect of number of related primes on brain activity, a multiple regression was applied for ambiguous and unambiguous trials separately with number of related primes as covariates (RR/RRa = 2, R1/R2/Rd/Rs = 1, UU/UUa = 0). Second, an SPM5 random-effects group analysis was performed by entering parameter estimates of the conditions (RR, R1, R2, UU, RRa, Rd, Rs, UUa) into a flexible factorial ANOVA. The main effect of ambiguity was modeled with an F-contrast. Third, we conducted post hoc conjunction analyses on the result of the regression for unambiguous trials (with number of related primes as a negative regressor) and the differential contrasts of the ANOVA that yielded activation in a comparable region in the posterior left inferior prefrontal cortex (pLIPC). All contrasts were corrected on a voxelwise threshold of P < 0.001. Hereby, a Monte Carlo simulation of the brain volume of the current study was conducted to establish an appropriate voxel contiguity threshold (Slotnick et al. 2003). Assuming an individual voxel type I error of P < 0.001, a cluster extent of 12 contiguous resampled voxels was indicated as necessary to correct for multiple voxel comparisons at P < 0.05. An inclusive mask of the main effect of ambiguity (P < 0.001, uncorrected) on all differential contrasts and conjunction analyses was used to restrict activation patterns to the global difference across ambiguity levels.

Anatomical labels for activated brain areas were computed using the SPM anatomy toolbox, which is based on anatomical probability maps of the brain reflecting the intersubject variability of selected brain areas in terms of location and extent (Eickhoff et al. 2005, 2006, 2007). Detailed anatomical maps were available of the inferior frontal gyrus (IFG; Brodmann areas [BAs] 44 and 45; Amunts et al. 1999), the inferior (Caspers et al. 2008) and superior parietal cortex (BAs 5 and 7; Scheperjans, Grefkes, et al. 2005; Scheperjans, Palomero-Gallagher, et al. 2005; Scheperjans et al. 2007), and the premotor cortex (BA 6; Geyer 2003). Activation clusters that did not match any probability map or the probability of whose were too small to get assigned to (<30%) were labeled on the macroanatomical level (e.g., left IFG [pars opercularis] and right rectal gyrus).

### Behavioral Analysis

Trials during which participants responded too slow (+2 SD), gave none or more than one response, or committed an error were excluded from further data analysis. An error was committed when subjects detected a prime–target relationship during unrelated trials and when no prime–target relationship was recognized during related trials. A 2 × 4 × 2 repeated-measures ANOVA with ambiguity (ambiguous, unambiguous), relation (RR, RU, UR, UU), and position (first, second) as within-subject factors was performed to identify significant effects of error rate and RT.

## Results

### Behavioral Results

On average, 3.08% (SD = 2.21) of the data had to be discarded due to misses, multiple responses, and outliers and 16.01% due to errors (unambiguous: 10.89% and ambiguous: 21.22%). Error rate and RT for each condition are listed in Table 2.

Table 2

RT and error rate for all 16 conditions recorded during fMRI in 15 healthy male adults (SDs are shown in parentheses)

 Unambiguous Ambiguous Type RT/ms Error (%) Type RT/ms Error (%) RR R1R2 788.93 (298.52) 0.56 (1.47) RRa RdRs 909.26 (315.37) 3.73 (5.34) R2R1 794.72 (290.22) 1.40 (2.59) RsRd 917.63 (320.75) 3.88 (4.52) RU R1U1 997.39 (319.04) 13.74 (11.03) Rd RdU1 955.37 (314.50) 11.85 (7.16) U1R1 1013.78 (304.05) 13.98 (8.69) U1Rd 1032.14 (323.63) 18.48 (14.06) UR R2U2 1007.63 (334.85) 10.47 (7.85) Rs RsU2 1137.00 (391.07) 32.74 (15.43) U2R2 1014.30 (252.84) 20.41 (12.85) U2Rs 1081.59 (314.24) 35.38 (15.59) UU U1U2 1340.44 (432.16) 17.21 (7.96) UUa U1U2a 1380.14 (499.18) 12.42 (12.49) U2U1 1359.18 (441.59) 9.35 (7.67) U2U1a 1333.80 (439.12) 8.82 (7.36)
 Unambiguous Ambiguous Type RT/ms Error (%) Type RT/ms Error (%) RR R1R2 788.93 (298.52) 0.56 (1.47) RRa RdRs 909.26 (315.37) 3.73 (5.34) R2R1 794.72 (290.22) 1.40 (2.59) RsRd 917.63 (320.75) 3.88 (4.52) RU R1U1 997.39 (319.04) 13.74 (11.03) Rd RdU1 955.37 (314.50) 11.85 (7.16) U1R1 1013.78 (304.05) 13.98 (8.69) U1Rd 1032.14 (323.63) 18.48 (14.06) UR R2U2 1007.63 (334.85) 10.47 (7.85) Rs RsU2 1137.00 (391.07) 32.74 (15.43) U2R2 1014.30 (252.84) 20.41 (12.85) U2Rs 1081.59 (314.24) 35.38 (15.59) UU U1U2 1340.44 (432.16) 17.21 (7.96) UUa U1U2a 1380.14 (499.18) 12.42 (12.49) U2U1 1359.18 (441.59) 9.35 (7.67) U2U1a 1333.80 (439.12) 8.82 (7.36)

Note: R, related; U, unrelated; a, ambiguous; d, prime addressing the dominant meaning of the homonym; s, prime addressing the subordinate meaning of the homonym.

#### Error Rate

The ANOVA for error rate revealed a main effect of ambiguity (F1,14 = 46.429, P < 0.001) and relation (F3,42 = 47.893, P < 0.001) and an ambiguity by relation (F3,42 = 14.344, P < 0.001), relation by position (F3,42 = 6.118, P < 0.005), and ambiguity by relation by position interaction (F3,42 = 4.240, P = 0.01). On average, participants committed more errors during the ambiguous trials.

Separate ANOVAs for each level of ambiguity confirmed the main effect of relation (F3,42 = 24.090, P < 0.001) and the relation by position interaction (F3,42 = 12.119, P < 0.001) for unambiguous targets. Participants were most accurate in double related trials (P < 0.001) and performed similarly across single and unrelated conditions (P = 1.0). Pairwise comparisons across conditions showed an effect of position for unrelated trials (P < 0.005) and trials with R2 (P < 0.005) but not for trials with R1 (P > 0.9) and double related trials (P > 0.3).

For ambiguous targets, the ANOVA revealed a main effect of relation only (F3,42 = 35.535, P < 0.001). Significantly more errors were committed for single related trials that addressed the subordinate meaning of the target compared with all other trials (P < 0.005). Participants were most accurate during double related trials, although the difference to unrelated trials was only marginal (P = 0.071).

Across ambiguity levels, pairwise comparisons revealed higher errors rates for subordinate-biasing trials relative to the respective unambiguous conditions (RsU vs. R2U: P < 0.001; URs vs. UR2: P = 0.006). The difference for ambiguous double related compared with unambiguous trials was significant (RdRs vs. R1R2: P = 0.049) or only marginal (RsRd vs. R2R1: P = 0.072), and no differences were observed for unrelated trials or those addressing the dominant meaning and their respective unambiguous conditions.

#### The RT

The ANOVA for RT revealed a main effect of ambiguity (F1,14 = 8.808, P = 0.01) and relation (F3,42 = 41.547, P < 0.001) and an ambiguity by relation interaction (F3,42 = 10.762, P < 0.001). On average, participants responded slower to ambiguous than unambiguous targets. No effect of position was observed (F1,14 = 0.890, P > 0.3), so that data were pooled accordingly for further RT and fMRI analyses.

Separate ANOVAs for each level of ambiguity replicated the main effect of relation (unambiguous: F3,42 = 48.334, P < 0.001; ambiguous: F3,42 = 31.171, P < 0.001). RT significantly increased (P < 0.005) from double to single to unrelated trials independent of ambiguity level. Single related trials did not differ for unambiguous targets (P = 1.0), but there was a trend toward slower RTs for subordinate-biasing compared with dominant-biasing contexts (P = 0.086). Pairwise comparisons between each condition across ambiguity level showed slower RTs for ambiguous targets in the double related condition (P < 0.001) and when the prime addressed the subordinate meaning in the single related condition (P = 0.054). Unrelated and dominant contexts were treated alike (P > 0.5).

### fMRI Results

#### Regression Analysis: Bottom-Up Support

The results of the regression analysis for unambiguous trials revealed increasing brain activation with increasing number of related primes in a bilateral frontoparietal network. This network comprised the angular gyrus (AG) and the rostromedial frontal cortex, which covered aspects of the rectal, orbitofrontal, superior frontal, and anterior cingulate gyrus (see Fig. 1 and Table 3). On the right hemisphere, the parietal activation reached into the middle occipital gyrus (MOG). Additional BOLD signal increases were seen in the middle temporal gyrus, medially in the middle cingulate cortex (mCC) and precuneus (BA 7), and a cluster spanning the left superior frontal gyrus (SFG) and middle frontal gyrus (MFG).

Table 3

Results of the regression analysis showing increases/decreases of brain activation with increasing number of related primes prior to target presentation for unambiguous and ambiguous targets

 Cluster extension Cl Unambiguous Cl Ambiguous Brain region Coordinates Z Brain region Coordinates Z Macroanatomy Microanatomy x y z Macroanatomy Microanatomy x y z Activation increase L/R Rostromedial frontal 311 recG 4 40 −20 5.47 199 mSFG −4 56 0 4.65 mSFG 8 64 20 4.96 SFG −24 60 8 4.45 mSFG −8 64 20 4.57 MFG −36 60 4 4.01 mOG 0 52 −8 4.51 ACC 4 40 −4 3.89 SFG −20 60 12 4.34 ACC −4 52 0 4.19 L Lateral frontal 89 SFG −20 40 44 4.57 116 SFG −12 36 52 4.98 MFG −44 12 52 3.90 MFG −44 16 52 4.21 L Inferior parietal 192 AG IPC (PGp/PGa) −44 −72 36 6.35 183 AG IPC (PGa/PFm) −56 −60 36 5.01 AG −44 −76 44 5.64 R Inferior parietal 51 AG IPC (PGp) 48 −72 36 5.35 56 AG IPC (PGp/PGa) 44 −68 48 4.52 MOG IPC (PGp) 48 −72 28 5.29 MOG IPC (PGp) 52 −68 28 3.83 L Temporal 17 MTG −60 −16 −24 3.57 12 ITG −64 −24 −20 3.84 L Medial parietal 31 mCC −8 −48 36 4.03 Precuneus SPL (BA 7) −4 −68 36 3.30 Activation decrease L Inferior frontal 22 Pars opercularis BA 44 −40 8 28 3.62 Pars triangularis −36 20 28 3.61 L/R Medial frontal 28 SMA −8 16 48 4.04 SMA BA 6 −4 12 52 3.79
 Cluster extension Cl Unambiguous Cl Ambiguous Brain region Coordinates Z Brain region Coordinates Z Macroanatomy Microanatomy x y z Macroanatomy Microanatomy x y z Activation increase L/R Rostromedial frontal 311 recG 4 40 −20 5.47 199 mSFG −4 56 0 4.65 mSFG 8 64 20 4.96 SFG −24 60 8 4.45 mSFG −8 64 20 4.57 MFG −36 60 4 4.01 mOG 0 52 −8 4.51 ACC 4 40 −4 3.89 SFG −20 60 12 4.34 ACC −4 52 0 4.19 L Lateral frontal 89 SFG −20 40 44 4.57 116 SFG −12 36 52 4.98 MFG −44 12 52 3.90 MFG −44 16 52 4.21 L Inferior parietal 192 AG IPC (PGp/PGa) −44 −72 36 6.35 183 AG IPC (PGa/PFm) −56 −60 36 5.01 AG −44 −76 44 5.64 R Inferior parietal 51 AG IPC (PGp) 48 −72 36 5.35 56 AG IPC (PGp/PGa) 44 −68 48 4.52 MOG IPC (PGp) 48 −72 28 5.29 MOG IPC (PGp) 52 −68 28 3.83 L Temporal 17 MTG −60 −16 −24 3.57 12 ITG −64 −24 −20 3.84 L Medial parietal 31 mCC −8 −48 36 4.03 Precuneus SPL (BA 7) −4 −68 36 3.30 Activation decrease L Inferior frontal 22 Pars opercularis BA 44 −40 8 28 3.62 Pars triangularis −36 20 28 3.61 L/R Medial frontal 28 SMA −8 16 48 4.04 SMA BA 6 −4 12 52 3.79

Note: L, left; R, right; ACC, anterior cingulate cortex; PFm, posterior aspect of the rostral part of the inferior parietal cortex; PGa/PGp, anterior/posterior aspect of the caudal inferior parietal cortex; ITG, inferior temporal gyrus; mOG, mid-orbital gyrus; mSFG, medial aspect of superior frontal gyrus; SPL, superior parietal lobule; MTG, middle temporal gyrus; recG, rectal gyrus; SMA, supplementary motor area. Coordinates are listed in MNI space. The significance level is given in Z values and cluster size (Cl) in number of voxels for P < 0.05 (corrected for multiple comparisons), cluster extent = 12 voxels.

Figure 1.

Results of the regression analysis for unambiguous and ambiguous trials. Brain activation increases (red) and decreases (blue) with the number of related primes preceding the target. Activation is corrected for multiple comparisons at P < 0.05 with a cluster extent of 12 voxels.

Figure 1.

Results of the regression analysis for unambiguous and ambiguous trials. Brain activation increases (red) and decreases (blue) with the number of related primes preceding the target. Activation is corrected for multiple comparisons at P < 0.05 with a cluster extent of 12 voxels.

The regression analysis for ambiguous trials revealed a similar increase of frontoparietal activation with increasing number of related primes. Areas included the rostromedial frontal cortex, covering the anterior cingulate, SFG and MFG, and the AG bilaterally (see Fig. 1 and Table 3). Again, the parietal activation reached into the MOG on the right hemisphere. Additional BOLD signal enhancements comprised the inferior temporal gyrus and a cluster covering the SFG and MFG on the left hemisphere.

#### Regression Analysis: Top-Down Control

The results of the regression analysis that identified top-down related brain structures were different for ambiguous and unambiguous trials. Increased BOLD responses were observed when the number of related primes decreased in the pLIPC (pars opercularis [BA 44] and triangularis) and the left pre-supplementary motor area (pre-SMA/BA 6) for unambiguous trials only (see Fig. 1 and Table 3). No corresponding linear increase of brain activation was observed when homonyms served as targets (see Fig. 1 and Table 3).

#### Effect of Ambiguity

An F-contrast of ambiguous relative to unambiguous trials across relatedness conditions revealed brain activation in anterior (∼BA 47) and pLIPC (BAs 44 and 45), the left precentral gyrus (precG), the right thalamus, and a cluster covering the right retrosplenial cortex (RSC; BA 30) and posterior cingulate cortex (pCC) and mCC (BA 5) (see Fig. 2 and Table 4A). Pairwise comparisons across the 3 major relatedness conditions (double, single, unrelated) showed that the latter medial parietal cluster and thalamus were significantly stronger and exclusively activated for unambiguous relative to ambiguous double related trials (see also Parameter estimates in Fig. 2 and Table 4B). No other brain area showed stronger BOLD signals for unambiguous compared with ambiguous conditions.

Table 4

(A) Main effect of ambiguity (F-contrast); differential brain activation ascribed to (B) unambiguous and (C) ambiguous trials across each of the 3 relatedness conditions (double, single, and unrelated) and (D) for single related conditions addressing the dominant and subordinate meaning separately

 H Brain area BA Coordinates Z Cl x y z A. Main effect of ambiguity L IFG (p. triangul.) 45 −48 24 28 5.27 178 L IFG (p. orbit.) −48 28 −4 4.94 L IFG (p. opercul.) 44 −52 16 20 4.82 L precG −48 8 40 3.45 R RSC 30 8 −52 20 4.01 75 R pCC 4 −48 28 3.93 R mCC/SPL 5 4 −36 44 3.81 R Thalamus 12 −32 4 4.22 14 B. Unambiguous Unrelated (UU > UUa) No region of statistical significance Single related (R1 + R2 > Rd + Rs) No region of statistical significance Double related (RR > RRa) R RSC 29 8 −48 8 4.41 67 R pCC 4 −48 24 4.14 R mCC 8 −40 40 3.93 R Thalamus 12 −32 4 5.00 14 C. Ambiguous Unrelated (UUa > UU) No region of statistical significance Single related (Rd + Rs > R1 + R2) L IFG (p. triangul.) 45 −48 28 16 4.23 40 L IFG (p. triangul.) 44 −44 12 24 3.89 Double-related (RRa > RR) L IFG (p. triangul.) 45 −44 20 28 4.07 32 L MFG −48 24 32 4.02 D. Dominant versus subordinate Dominant (Rd > R1) No region of statistical significance Subordinate (Rs > R2) L IFG (p. triangul.) 45/44 −48 24 24 5.17 145 L IFG (p. orbit.) −48 28 −4 5.05 Dominant > subordinate No region of statistical significance Subordinate > dominant L IFG (p. orbit.) −52 32 −4 5.32 66 L IFG (p. triangul.) 45/44 −52 20 24 4.17 47 L precG −48 8 40 3.58
 H Brain area BA Coordinates Z Cl x y z A. Main effect of ambiguity L IFG (p. triangul.) 45 −48 24 28 5.27 178 L IFG (p. orbit.) −48 28 −4 4.94 L IFG (p. opercul.) 44 −52 16 20 4.82 L precG −48 8 40 3.45 R RSC 30 8 −52 20 4.01 75 R pCC 4 −48 28 3.93 R mCC/SPL 5 4 −36 44 3.81 R Thalamus 12 −32 4 4.22 14 B. Unambiguous Unrelated (UU > UUa) No region of statistical significance Single related (R1 + R2 > Rd + Rs) No region of statistical significance Double related (RR > RRa) R RSC 29 8 −48 8 4.41 67 R pCC 4 −48 24 4.14 R mCC 8 −40 40 3.93 R Thalamus 12 −32 4 5.00 14 C. Ambiguous Unrelated (UUa > UU) No region of statistical significance Single related (Rd + Rs > R1 + R2) L IFG (p. triangul.) 45 −48 28 16 4.23 40 L IFG (p. triangul.) 44 −44 12 24 3.89 Double-related (RRa > RR) L IFG (p. triangul.) 45 −44 20 28 4.07 32 L MFG −48 24 32 4.02 D. Dominant versus subordinate Dominant (Rd > R1) No region of statistical significance Subordinate (Rs > R2) L IFG (p. triangul.) 45/44 −48 24 24 5.17 145 L IFG (p. orbit.) −48 28 −4 5.05 Dominant > subordinate No region of statistical significance Subordinate > dominant L IFG (p. orbit.) −52 32 −4 5.32 66 L IFG (p. triangul.) 45/44 −52 20 24 4.17 47 L precG −48 8 40 3.58

Note: H, hemisphere; L, left; R, right; p. opercul., pars opercularis; p. orbit., pars orbitalis; p. triangul., pars triangularis; SPL, superior parietal lobule. Coordinates are listed in MNI space. The significance level is given in Z values and cluster size (Cl) in number of voxels for P < 0.05 (corrected for multiple comparisons), cluster extent = 12 voxels.

Figure 2.

Top panel: Brain activation reflecting the main effect of ambiguity. Peak activation of the inferior frontal and medial parietal cluster is displayed on coronal and sagittal slices of the Anatomy toolbox template. Bottom panel: Comparisons between ambiguous and unambiguous trials across relatedness condition. Left: Stronger LIPC activation was observed for ambiguous compared with unambiguous trials in the double related (RRa > RR) and single related condition ([Rd + Rs] > [R1 + R2]), in single related trials addressing the subordinate meaning compared with unambiguous single related trials (Rs > R2) and compared with trials addressing the dominant meaning ([Rs > R2] > [Rd > R1]). Right: Stronger activation was observed for unambiguous compared with ambiguous trials in the double related condition only (RR > RRa). Contrasts with no differential activation are not shown. Coronal and sagittal slices display peak activation. Contrast estimates are provided for regions marked by an asterisk. R, related; U, unrelated; a, ambiguous; Rd/Rs, trials addressing the dominant/subordinate meaning of the homonym. All differential contrasts are masked by the main effect of ambiguity at P < 0.001 (uncorrected). Activation is corrected for multiple comparisons at P < 0.05 with a cluster extent of 12 voxels.

Figure 2.

Top panel: Brain activation reflecting the main effect of ambiguity. Peak activation of the inferior frontal and medial parietal cluster is displayed on coronal and sagittal slices of the Anatomy toolbox template. Bottom panel: Comparisons between ambiguous and unambiguous trials across relatedness condition. Left: Stronger LIPC activation was observed for ambiguous compared with unambiguous trials in the double related (RRa > RR) and single related condition ([Rd + Rs] > [R1 + R2]), in single related trials addressing the subordinate meaning compared with unambiguous single related trials (Rs > R2) and compared with trials addressing the dominant meaning ([Rs > R2] > [Rd > R1]). Right: Stronger activation was observed for unambiguous compared with ambiguous trials in the double related condition only (RR > RRa). Contrasts with no differential activation are not shown. Coronal and sagittal slices display peak activation. Contrast estimates are provided for regions marked by an asterisk. R, related; U, unrelated; a, ambiguous; Rd/Rs, trials addressing the dominant/subordinate meaning of the homonym. All differential contrasts are masked by the main effect of ambiguity at P < 0.001 (uncorrected). Activation is corrected for multiple comparisons at P < 0.05 with a cluster extent of 12 voxels.

Brain activation in the left prefrontal cortex was significantly stronger for ambiguous compared to unambiguous targets, covering posterior parts of the LIPC (BAs 44 and 45) for single related trials and a cluster covering BA 45 and the adjacent MFG for double related trials (see Fig. 2 and Table 4C). No differential activation was found for the unrelated condition across ambiguity levels.

#### Effect of Dominant- and Subordinate-Biasing Contexts

A more detailed analysis of the ambiguous single related trials, classifying them into those with subordinate and dominant-biasing contexts, revealed a pronounced left frontal activation pattern for subordinate trials only (see Fig. 2 and Table 4D). This activation spanned the pLIPC (pars triangularis, BA 45/44) and the anterior left inferior prefrontal cortex (aLIPC) (∼BA 47) relative to the unambiguous single related condition. In direct comparison to dominant single related trials (relative to the unambiguous counterparts), the LIPC (BA 45/44, ∼BA 47) remained significant for subordinate-biasing contexts, extending dorsally into the precG (see Fig. 2 and Table 4D).

#### Results of the Conjunction Analyses

To determine whether the pLIPC activation that decreased linearly with the number of related primes for unambiguous trials (see regression analysis) overlapped with the one observed during certain ambiguous compared with their unambiguous conditions (i.e., double related, single related, subordinate), we conducted 3 independent conjunction analyses. The activation cluster for ambiguous vs. unambiguous double related trials was not congruent with the frontal activation identified by the regression analysis. Neither was an overlap observed for the single related ambiguous trials compared with their unambiguous conditions. In contrast, we found common activation in the pars opercularis/triangularis (∼BA 44/45) for subordinate-biasing contexts (vs. the respective unambiguous condition) and the results of the regression analysis (peak activation: x = −40, y = 8, z = 28, Z = 3.72, cluster size = 14 voxels; local maxima: x = −40, y = 20, z = 24, Z = 3.44) (Fig. 3).

Figure 3.

Brain activation common to the regression analysis with the number of related primes as a negative regressor and the differential contrast comparing subordinate ambiguous trials (Rs) to the respective single related unambiguous trials (R2). (A) Peak activation and (B) activation at the local maxima are superimposed onto the cytoarchitectonic probability maps from the Anatomy toolbox. Activation is corrected for multiple comparisons at P < 0.05 with a cluster extent of 12 voxels.

Figure 3.

Brain activation common to the regression analysis with the number of related primes as a negative regressor and the differential contrast comparing subordinate ambiguous trials (Rs) to the respective single related unambiguous trials (R2). (A) Peak activation and (B) activation at the local maxima are superimposed onto the cytoarchitectonic probability maps from the Anatomy toolbox. Activation is corrected for multiple comparisons at P < 0.05 with a cluster extent of 12 voxels.

## Discussion

In the current study, we investigated the impact of prior semantic information on meaning retrieval using multiple primes. We were able to show that target processing was linearly dependent on the amount of prior semantic material and that 2 different retrieval networks (i.e., bottom-up, top-down related) responded to such manipulations. We induced bottom-up support by increasing the number of related primes prior to target presentation and observed increased activation in a rostromedial–frontoparietal network (see regression analysis). This modulation was independent of ambiguity level, thus similar for ambiguous and unambiguous targets. Interestingly, when bottom-up support was supposed to be maximal (during unambiguous double related trials), the RSC and pCC were selectively engaged (see parameter estimates Fig. 2). Parallel to an increase of bottom-up support, top-down control was postulated to decrease. Two brain areas, the pLIPC and pre-SMA, showed a gradual reduction of brain activation for unambiguous targets only (see regression analysis). Top-down control was also manipulated by presenting homonyms as targets. Comparisons between ambiguous and unambiguous conditions revealed that specific aspects of the LIPC were distinctively engaged during ambiguous trials. The pLIPC responded when both meanings of the homonym were primed, whereas the aLIPC was additionally recruited for subordinate-biasing contexts. Finally, the post hoc conjunction analysis confirmed an overlap between pLIPC activation during subordinate trials and the regression analysis.

On the basis of our fMRI data, we will propose 2 retrieval networks that mediate bottom-up and top-down related retrieval processes. The results of all 3 statistical approaches, the regression analysis, the direct comparisons across ambiguity level, and the conjunction, are used to specify the function of each individual brain structure in these networks.

### Bottom-Up Support during Meaning Retrieval: The AG and Rostromedial Prefrontal Cortex

From previous multiple prime studies and our own behavioral data, we know that participants’ judgments at the final word are extremely fast when both primes converge onto the target concept (lion–stripe–tiger). This decision process does not allow for time-consuming top-down control but seems to work on automatic, stimulus-driven mechanisms instead. A distinct network of brain regions, primarily including the bilateral AG and rostromedial prefrontal cortex (rmPC), responded most when judgments were carried out rapidly.

Magnetoencephalography investigations were able to track the temporal dynamics of visual word/picture processing (for a review on silent word reading, see Salmelin and Kujala 2006; for word production, see Indefrey and Levelt 2004; e.g., Vihla et al. 2006; Salmelin 2007), generating predictions about brain areas involved in early, bottom-up support. Those studies identified a posterior-to-anterior processing stream that emerged in the occipital sensory cortex and proceeded along the ventrotemporal region into bilateral parietal/temporal cortex mediating prelexical to higher order lexicosemantic and phonological functions, respectively, before entering the lateral prefrontal and premotor cortex for (optional) speech output preparation. At around 400-ms poststimulus, brain regions in and around the posterior superior temporal cortex have been linked to the ease of contextual integration, reflected in amplitude changes of electric or magnetic potentials (N400, N400m) in response to semantic incongruities (e.g., Kutas and Hillyard 1980; Halgren et al. 2002; Pylkkänen and Marantz 2003; Salmelin and Kujala 2006). In contrast, a later electrophysiological potential (N450) over the dorsolateral prefrontal/anterior cingulate cortex could be linked to increased demands in cognitive control or conflict monitoring during classic executive tasks (e.g., the Stroop task; West et al. 2005; Larson et al. 2008). Meaning retrieval that is based on a stimulus-driven route might thus rely on posterior language-related areas, which is in agreement with the AG activation we associated with bottom-up support in the current study.

The AG is a core component of semantic processing (for reviews, see Price 2000; Démonet et al. 2005; e.g., Roskies et al. 2001; Binder et al. 2003, 2005; Rissman et al. 2003; Scott et al. 2003). Recent neuroimaging investigations have attributed activation of the AG to higher order combinatorial semantics, that is, the integration of semantic input into an ongoing sentential or narrative context, rather than lexicosemantic processing per se. Evidence was provided by studies that compared sentences with unrelated word lists or short text excerpts to passages of pseudo-sentences (Ferstl and von Cramon 2002; Humphries et al. 2005, 2006, 2007; Xu et al. 2005; Sieboerger et al. 2007). The AG also responded when integration was facilitated by presenting semantically high- versus low-predictable sentences under adverse listening conditions (Obleser et al. 2007). Even in situations of minimal contextual embedding (e.g., word pairs), integration efforts and AG response were manipulated. In a recent semantic priming study, Kuperberg et al. (2008) reported BOLD enhancements in the AG for related compared with unrelated word pairs. Consistent with the integrative approach, a stronger BOLD response of the AG was observed in our investigation when prime and target were related and integration of the target into its preceding context was successful.

The ability to integrate incoming linguistic material into a coherent context is likely to require further cognitive supportive mechanisms, for example, attention. In order to integrate 2 concepts, participants have to specifically attend toward common semantic features between primes and target (cf., Kuperberg et al. 2008) and selectively focus on these externally provided cues to allow for rapid decisions (cf., Obleser et al. 2007). The inferior parietal lobe itself has been linked to the direct and automatic allocation of attention especially when retrieval is triggered by salient environmental memory cues, for example, strong prime–target associations (for a review, see Ciramelli et al. 2008). A similar function can be ascribed to the medial prefrontal cortex.

The diversity of cognitive tasks during which activation was observed in the rmPC implies that this brain region subserves domain-independent cognitive functions (for a review, see Ramnani and Owen 2004). In a recent meta-analysis, activation of the rmPC, which corresponds to medial BA 10, has been associated with shorter response latencies during experimental compared with control conditions across a variety of cognitive domains, including language, episodic memory, or mentalizing (for a meta-analysis, see Gilbert, Spengler, Simons, Steel, et al. 2006). The authors hypothesized that activation of the rmPC reflected attention toward external (i.e., stimulus oriented) cues rather than shifting attention toward internal (i.e., stimulus independent) or self-referential thought processes (Burgess et al. 2003, 2007; Gilbert, Simons, et al. 2006; Gilbert, Spengler, Simons, Frith, et al. 2006). Consistent with this approach, recruitment of the rmPC was apparent when external cues (i.e., primes) sufficiently specified or prepared an upcoming event (i.e., target) compared with underspecified environmental cues in our study or in a study from a different cognitive domain (see for target detection, Small et al. 2003). Our results are in line with the hypothesis that the rmPC acts as a neural component of a stimulus-driven attention network, which allows for fast decisions on stimuli that are embedded in highly specified contextual environments.

### The Role of the Precuneus and Posterior Cingulate Gyrus in Bottom-Up Support

The posterior cingulate gyrus (i.e., RSC and pCC) was distinctively engaged during the unambiguous double related condition (lion–stripe–tiger); the only condition where the contextual environment was completely coherent and bottom-up support was hypothesized to be maximal. The bordering precuneus further showed a linear response to increased bottom-up support (see regression analysis). Neuroimaging studies of story and sentence comprehension have associated precuneus and posterior cingulate activity with successful narrative integration and the establishment of coherence across sentences (Ferstl and von Cramon 2001, 2002; Xu et al. 2005; Speer et al. 2007). Similarly, these structures have been linked to the online integration of sentence-final words into a coherent context (Kircher et al. 2001; Mestres-Missé et al. 2008). Again, coherence building demands additional cognitive resources. Incoming semantic information needs to be evaluated and retained if a common feature can be extracted and a coherent situation built up. Both, the precuneus and pCC have been postulated to be part of a larger parietal working memory buffer that accumulates information for subsequent decision-making processes and evaluates the input being perceived (for reviews, see Vogt et al. 1992; Wagner et al. 2005). As soon as semantic information is encountered that cannot be linked to previous material held in working memory, activation decreases. In our study, neither during single related, unrelated, nor ambiguous double related trials, such a connection can be established. The parameter estimates of the posterior cingulate–precuneus cluster support a selective function of these brain structures for maintaining coherent contexts.

In sum, the bottom-up network that we identified by increasing the number of related primes prior to target presentation is highly distributed and includes frontal and parietal brain structures. The central component of this network seems to be the AG, activity of which increased with successful semantic integration. It was accompanied by cognitive–supportive mechanisms residing in medial frontal and parietal areas such as directing attention toward external stimuli and maintaining information for coherence building, respectively.

### Top-Down Control during Meaning Retrieval: pLIPC

The current study demonstrated a key role of the pLIPC in the top-down related recovery of meaning. pLIPC activation levels increased during several adverse retrieval conditions: when no preexisting associations facilitated the search process (i.e., during unrelated trials), during the retrieval of low-frequent target meanings (i.e., during subordinate trials) and when several candidate responses were recovered (i.e., during ambiguous double related trials). Analogue to our data, a distinct upregulation of pLIPC activity has been reported for unprimed relative to primed target words (Kotz et al. 2002), for low- versus high-frequent targets despite similar associative strength (Chee et al. 2002) and under conditions, where retrieval demands increased (Badre et al. 2005). The overlap of pLIPC activation from the regression analysis and the subordinate trials further emphasizes a highly flexible and adaptive behavior of this structure (BA 44/45) during situations that demand controlled access to word meaning. We assume that the pLIPC supports the exhaustive and strategic search for appropriate semantic concepts when automatic stimulus-driven support is insufficient, whether this is achieved by manipulation of prior semantic context or lexicosemantic properties of the target word (e.g., meaning frequency).

In contrast to unambiguous trials, BOLD activity in the pLIPC or pre-SMA was not modulated by multiple primes when homonyms served as targets. This result was not unexpected considering that access to a specific concept of a homonym is not dependent on prior contextual information alone but also on its lexicosemantic attributes. Accordingly, activation of the 2 brain regions that responded stronger to ambiguous versus unambiguous trials (i.e., aLIPC and pLIPC) was not linearly dependent on the number of related primes but particularly sensitive to subordinate-biasing trials.

### Top-Down Control during Meaning Retrieval: aLIPC

Due to its exclusive activation to subordinate-biasing trials, the function of the aLIPC seems to be more specific. We hypothesize that activity of this structure reflects the resolution of competition among alternative responses during semantic retrieval. In our study, the subordinate-biasing trials (dance–clock–ball) encouraged the retrieval of the weaker meaning of the homonym, whereas the coactivated dominant meaning needed to be suppressed. Increased error rates and RTs complemented this effect. This observation is supported by a recent investigation, in which the disambiguation of a biased homonym to the subordinate meaning resulted in aLIPC hyperactivity compared with balanced homonyms (Mason and Just 2007). The same brain structure was identified when competition increased among unambiguous concepts of similar word frequency, arguing against a low-frequency effect (Moss et al. 2005). The authors reported aLIPC activation during picture naming when a semantically related meaning was retrieved earlier during the experiment based on verbal descriptions of the concept.

Although the ambiguous double related condition (game–dance–ball) also addressed both concepts of the homonym, the selection and subsequent inhibition of the competing subordinate concept seemed less demanding (as indexed by faster RT for double than subordinate trials). Unlike subordinate-biasing trials, the ambiguous double related condition failed to evoke aLIPC activation compared with the unambiguous control condition. The data support hybrid models of ambiguity resolution, which claim an inherent preference for the high-frequent meaning of biased homonyms but enable the subordinate concept to compete in subordinate-biasing contexts for meaning selection (e.g., Rayner and Frazier 1989). In contexts, however, in which both meanings are primed, ambiguity is resolved toward the high-frequent concept because of additional contextual input onto the preactivation level of the preferred meaning.

### Functional Specialization within the Left Prefrontal Cortex

The controlled retrieval of appropriate verbal responses and its optional, subsequent selection from a set of competitors are the 2 most prominent and controversially debated functions ascribed to the LIPC. Most researchers agree, though, on a functional dissociation of the LIPC and that retrieval and selection processes are mediated by anterior and posterior portions of the LIPC, respectively (Wagner et al. 2001; Badre et al. 2005; Dobbins and Wagner 2005; Gold et al. 2006). Our data contradict this dichotomy. Although the study was not primarily designed to distinguish between controlled retrieval and selection demands, the ambiguity data provide convincing evidence for a role of the aLIPC in resolving competition during the recovery of meaning. Former investigations manipulated selection/competition demands by directing attention toward specific features of a concept while task-irrelevant associations needed to be suppressed (e.g., focusing on color or form vs. overall similarity, see Thompson-Schill et al. 1997). A more direct approach to address selection/competition demands can be achieved by introducing competition between alternative concepts rather than between certain features of a single concept. Moss et al. (2005) investigated competition on the concept-level and reported aLIPC activation. We presented biased homonyms, which required a context-dependent meaning selection, and observed activation in the same structure.

Our data also strongly imply a function of the pLIPC in the controlled recovery of semantic concepts independent of selection processes. A standard procedure to assess controlled retrieval involves the manipulation of the associative strength between cue and target concept (e.g., Wagner et al. 2001; Badre et al. 2005). We gradually increased the influence of facilitating cue-target associations in the unambiguous condition and observed a reduction of pLIPC activation. The same region was observed using a different analysis (ANOVA) and revealed that the pLIPC does not only respond to a decrease in relevant associations but also to conditions where other processes interfered with automatic knowledge retrieval (i.e., when the low-frequent meaning of a homonym was addressed).

Alternatively, the current findings can be integrated into a more universal model of cognitive control that is based on the complexity or abstractness of the representations that compete during the decision process (Badre and D'Esposito 2007; Badre 2008). The authors propose a hierarchical organization within the prefrontal cortex whereby more complex/abstract representations are processed in progressively anterior regions, ranging from the premotor cortex to the aLIPC/frontopolar cortex (Badre 2008). According to this model, decisions that are influenced by the current temporal/contextual environment demand the most complex form of control (i.e., contextual control) and are thus located most anterior in the prefrontal cortex. In contrast, decisions for which the current situation is irrelevant and selection is based on sensory rather than current “episodic” input reflect a lower level of abstraction in this hierarchy (e.g., on the feature level), yielding activation in more posterior regions. The relatedness judgment task in our study can be modeled as a decision process that either demands contextual/episodic control (for ambiguous targets) or can succeed without it (for unambiguous targets). For unambiguous targets, the response is independent of the current context, that is, “stripes–tiger” will generally be classed as being related, whereas “bread–tiger” will be judged as unrelated. In contrast, the relatedness response between a prime and a homonym (e.g., “dance–ball”) depends on the current contextual/temporal constraints the homonym is being placed in, particularly when these constraints deviate from the default (i.e., dominant) state. The relatedness judgment is only positive if “ball” is considered in its subordinate context but negative if the homonym is linked to its dominant meaning (e.g., referring to football). Consistent with this interpretation, we observed aLIPC activation for subordinate trials only and failed to show any context-sensitive BOLD response during unambiguous or dominant-biasing trials.

Apart from ambiguity resolution, the model further predicts that episodic control should be relevant for decision processes that have been manipulated by previous learning or priming procedures. Hereby, competition arises between the primed/learned context and the unaffected, default situation and not, as previously, between subordinate and dominant contexts. For example, Moss (2005) reported aLIPC activation during naming of a target concept (e.g., shark) that had been primed on a previous trial by a competing, semantically related concept (e.g., whale) compared with an unprimed target. The hierarchical control model might also resolve the discrepancy in brain activation between our results and former studies that manipulated selection/competition demands (Badre et al. 2005; Thompson-Schill et al. 1997). Such previous investigations evoked high selection demands by directing attention toward specific features of cue-target pairs. The decisions did not require contextual control and thus produced brain activation in the posterior prefrontal cortex. However, it remains debatable in how far this approach can explain aLIPC activation in studies that manipulated cue-target associations (Wagner et al. 2001; Badre et al. 2005) during meaning retrieval, a condition that does not seem to involve contextual control.

The current study revealed that meaning retrieval in context is supported by 2 neural networks, which were spatially distinct and held different functional properties. The frontoparietal network seems to support bottom-up related retrieval processes independent of ambiguity level and is based on brain regions that have been linked to semantic integration and universal attention and working memory functions. Top-down control appears to be more restricted to structures of executive semantic processing and particularly sensitive to different kinds of retrieval manipulations (i.e., controlled vs. automatic, competition vs. unambiguousness, episodic vs. sensory control). For a successful recovery of meaning, a flexible adjustment of top-down and bottom-up regulation is required, which can be manipulated by the contextual environment.

## Funding

The German Research Foundation [1328 to C.W.].

We thank Simon Eickhoff for valuable comments on the fMRI data analysis during the revision process. Conflict of Interest: None declared.

## References

Amaro
E
Jr
Barker
GJ
Brain Cogn
,
2006
, vol.
60
(pg.
220
-
232
)
Amunts
K
Schleicher
A
Burgel
U
Mohlberg
H
Uylings
HB
Zilles
K
Broca's region revisited: cytoarchitecture and intersubject variability
J Comp Neurol
,
1999
, vol.
412
(pg.
319
-
341
)
Annett
M
Classification of hand preference by association analysis
Br J Psychiatry
,
1970
, vol.
61
(pg.
303
-
321
)
Baayen
RH
Piepenbrock
R
van Rijn
H
The CELEX lexical database (CD-ROM)
,
1993
Linguistic Data Consortium, University of Pennsylvania
D
Cognitive control, hierarchy, and the rostro-caudal organization of the frontal lobes
Trends Cogn Sci
,
2008
, vol.
12
(pg.
193
-
200
)
D
D'Esposito
M
Functional magnetic resonance imaging evidence for a hierarchical organization of the prefrontal cortex
J Cogn Neurosci
,
2007
, vol.
19
(pg.
2082
-
2099
)
D
Poldrack
RA
Paré-Blagoev
EJ
Insler
RZ
Wagner
Dissociable controlled retrieval and generalized selection mechanisms in ventrolateral prefrontal cortex
Neuron
,
2005
, vol.
47
(pg.
907
-
918
)
Balota
DA
Paul
ST
Summation of activation: evidence from multiple primes that converge and diverge within semantic memory
J Exp Psychol
,
1996
, vol.
22
(pg.
827
-
845
)
Barsalou
LW
Simmons
WK
Barbey
AK
Wilson
CD
Grounding conceptual knowledge in modality-specific systems
Trends Cogn Sci
,
2003
, vol.
7
(pg.
84
-
91
)
Beeman
M
Friedman
RB
Grafman
J
Perez
E
Diamond
S
Lindsay
MB
Summation priming and coarse semantic coding in the right hemisphere
J Cogn Neurosci
,
1994
, vol.
6
(pg.
26
-
45
)
Binder
JR
McKiernan
KA
Parsons
ME
Westbury
CF
Possing
ET
Kaufman
JN
Buchanan
L
Neural correlates of lexical access during visual word recognition
J Cogn Neurosci
,
2003
, vol.
15
(pg.
372
-
393
)
Binder
JR
Westbury
CF
McKiernan
KA
Possing
ET
Medler
DA
Distinct brain systems for processing concrete and abstract concepts
J Cogn Neurosci
,
2005
, vol.
17
(pg.
905
-
917
)
Bookheimer
S
Functional MRI of language: new approaches to understanding cortical organization of semantic processing
Annu Rev Neurosci
,
2002
, vol.
25
(pg.
151
-
188
)
Burgess
PW
Dumontheil
I
Gilbert
SJ
The gateway hypothesis of rostral prefrontal cortex (area 10) function
Trends Cogn Sci
,
2007
, vol.
11
(pg.
290
-
298
)
Burgess
PW
Scott
SK
Frith
CD
The role of the rostral frontal cortex (area 10) in prospective memory: a lateral versus medial dissociation
Neuropsychologia
,
2003
, vol.
41
(pg.
906
-
918
)
Caspers
S
Eickhoff
SB
Geyer
S
Scheperjans
F
Mohlberg
H
Zilles
K
Amunts
K
The human inferior parietal lobule in stereotaxic space
Brain Struct Funct
,
2008
, vol.
212
(pg.
481
-
495
)
Chee
MWL
Hon
NHH
Caplan
D
Lee
HL
Goh
J
Frequency of concrete words modulates prefrontal activation during semantic judgments
Neuroimage
,
2002
, vol.
16
(pg.
259
-
268
)
Chwilla
DJ
Kolk
HHJ
Event-related potential and reaction time evidence for inhibition between alternative meanings of ambiguous words
Brain Lang
,
2003
, vol.
86
(pg.
167
-
192
)
Ciramelli
E
CL
Moscovitch
M
Top-down and bottom-up attention to memory: a hypothesis (AtoM) on the role of the posterior parietal cortex in memory retrieval
Neuropsychologia
,
2008
, vol.
46
(pg.
1828
-
1851
)
Collins
AM
Loftus
EF
A spreading-activation theory of semantic processing
Psychol Rev
,
1975
, vol.
82
(pg.
407
-
428
)
Copland
DA
de Zubicaray
GI
McMahon
K
Wilson
SJ
Eastburn
M
Chenery
HJ
Brain activation during automatic semantic priming revealed by event-related functional magnetic imaging
Neuroimage
,
2003
, vol.
20
(pg.
302
-
310
)
Davis
MH
Coleman
MR
Absalom
AR
Rodd
JM
Johnsrude
IS
Matta
BF
Owen
AM
Menon
DK
Dissociating speech perception and comprehension at reduced levels of awareness
,
2007
, vol.
104
(pg.
16032
-
16037
)
Démonet
J-F
Thierry
G
Cardebat
D
Renewal of the neurophysiology of language: functional neuroimaging
Physiol Rev
,
2005
, vol.
85
(pg.
49
-
95
)
Dobbins
IG
Wagner
Domain-general and domain-sensitive prefrontal mechanisms for recollecting events and detecting novelty
Cereb Cortex
,
2005
, vol.
15
(pg.
1768
-
1778
)
Duffy
SA
Morris
RK
Rayner
K
Lexical ambiguity and fixation times in reading
J Mem Lang
,
1988
, vol.
27
(pg.
429
-
446
)
Eickhoff
SB
Heim
S
Zilles
K
Amunts
K
Testing anatomically specified hypotheses in functional imaging using cytoarchitectonic maps
Neuroimage
,
2006
, vol.
32
(pg.
570
-
582
)
Eickhoff
SB
Paus
T
Caspers
S
Grosbras
M-H
Evans
AC
Zilles
K
Amunts
K
Assignment of functional activations to probabilistic cytoarchitectonic areas revisited
Neuroimage
,
2007
, vol.
36
(pg.
511
-
521
)
Eickhoff
SB
Stephan
KE
Mohlberg
H
Grefkes
C
Fink
GR
Amunts
K
Zilles
K
A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data
Neuroimage
,
2005
, vol.
25
(pg.
1325
-
1335
)
Faust
M
Lavidor
M
Semantically convergent and semantically divergent priming in the cerebral hemispheres: lexical decision and semantic judgment
Brain Res Cogn Brain Res
,
2003
, vol.
17
(pg.
585
-
597
)
Ferstl
EC
von Cramon
DY
The role of coherence and cohesion in text comprehension: an event-related fMRI study
Brain Res Cogn Brain Res
,
2001
, vol.
11
(pg.
325
-
340
)
Ferstl
EC
von Cramon
DY
What does the frontomedian cortex contribute to language processing: coherence or theory of mind?
Neuroimage
,
2002
, vol.
17
(pg.
1599
-
1612
)
Friston
KJ
Fletcher
P
Josephs
O
Holmes
A
Rugg
MD
Turner
R
Event-related fMRI: characterizing differential responses
Neuroimage
,
1998
, vol.
7
(pg.
30
-
40
)
Gennari
SP
MacDonald
MC
Postle
BR
Seidenberg
MS
Context-dependent interpretation of words: evidence for interactive neural processes
Neuroimage
,
2007
, vol.
35
(pg.
1278
-
1286
)
Geyer
S
The microstructural border between the motor and the cognitive domain in the human cerebral cortex
,
2003
Wien, Austria
Springer
Gilbert
SJ
Simons
JS
Frith
CD
Burgess
PW
Performance-related activity in medial rostral prefrontal cortex (area 10) during low-demand tasks
J Exp Psychol Hum Percept Perform
,
2006
, vol.
32
(pg.
45
-
58
)
Gilbert
SJ
Spengler
S
Simons
JS
Frith
CD
Burgess
PW
Differential functions of lateral and medial rostral prefrontal cortex (area 10) revealed by brain-behavior associations
Cereb Cortex
,
2006
, vol.
16
(pg.
1783
-
1789
)
Gilbert
SJ
Spengler
S
Simons
JS
Steele
JD
Lawrie
SM
Frith
CD
Burgess
PW
Functional specialization within rostral prefrontal cortex (area 10): a meta-analysis
J Cogn Neurosci
,
2006
, vol.
18
(pg.
932
-
948
)
Giora
R
On the priority of salient meanings: studies of literal and figurative language
J Pragm
,
1999
, vol.
31
(pg.
919
-
929
)
Gold
BT
Balota
DA
Jones
SJ
Powell
DK
Smith
CD
Andersen
AH
Dissociation of automatic and strategic lexical-semantics: functional magnetic resonance imaging evidence for differing roles of multiple frontotemporal regions
J Neurosci
,
2006
, vol.
26
(pg.
6523
-
6532
)
Gold
BT
Buckner
RL
Common prefrontal regions coactivate with dissociable posterior regions during controlled semantic and phonological tasks
Neuron
,
2002
, vol.
35
(pg.
803
-
812
)
Halgren
E
Dhond
RP
Christensen
N
van Petten
C
Marinkovic
K
Lewine
JD
Dale
AM
N400-like magnetoencephalography response modulated by semantic context, word frequency and lexical class in sentence
Neuroimage
,
2002
, vol.
17
(pg.
1101
-
1116
)
Humphries
C
Binder
JR
Medler
DA
Liebenthal
E
Syntactic and semantic modulation of neural activity during auditory sentence comprehension
J Cogn Neurosci
,
2006
, vol.
18
(pg.
665
-
679
)
Humphries
C
Binder
JR
Medler
DA
Liebenthal
E
Time course of semantic processes during sentence comprehension: an fMRI study
Neuroimage
,
2007
, vol.
36
(pg.
924
-
932
)
Humphries
C
Swinney
D
Love
T
Hickok
G
Response of anterior temporal cortex to syntactic and prosodic manipulations during sentence processing
Hum Brain Mapp
,
2005
, vol.
26
(pg.
128
-
138
)
Indefrey
P
Levelt
WJM
The spatial and temporal signatures of word production components
Cognition
,
2004
, vol.
92
(pg.
101
-
44
)
Jefferies
E
Lambon Ralph
MA
Semantic impairments in stroke aphasia versus semantic dementia: a case-series comparison
Brain
,
2006
, vol.
129
(pg.
2132
-
2147
)
P
Federmeier
KD
Multiple priming of lexically ambiguous and unambiguous targets in the cerebral hemispheres: the coarse coding hypothesis revisited
Brain Res
,
2007
, vol.
1153
(pg.
144
-
157
)
Kircher
TT
Brammer
M
Tous Andreu
N
Williams
SC
McGuire
PK
Engagement of right temporal cortex during processing of linguistic context
Neuropsychologia
,
2001
, vol.
39
(pg.
798
-
809
)
Kotz
SA
Cappa
SF
von Cramon
DY
Friederici
Modulation of the lexical-semantic network by auditory semantic priming: an event-related functional MRI study
Neuroimage
,
2002
, vol.
17
(pg.
1761
-
1772
)
Kuperberg
GR
Lakshmanan
BM
Greve
DN
West
WC
Task and semantic relationship influence both the polarity and localization of hemodynamic modulation during lexico-semantic processing
Hum Brain Mapp
,
2008
, vol.
29
(pg.
544
-
561
)
Kutas
M
Hillyard
SA
Reading senseless sentences: brain potentials reflect semantic incongruity
Science
,
1980
, vol.
207
(pg.
203
-
205
)
Larson
MJ
Kaufman
DAS
Perlstein
WM
Neuropsychologia
,
2008

Available online November 21, 2008, doi:10.1016/j.neuropsychologia.2008.11.013
Lehrl
S
Triebig
G
Fischer
B
Multiple-choice vocabulary test MWT as a valid and short test to estimate premorbid intelligence
Acta Neurol Scand
,
1995
, vol.
91
(pg.
335
-
345
)
Martin
A
The representation of object concepts in the brain
Annu Rev Psychol
,
2007
, vol.
58
(pg.
25
-
45
)
Martin
C
Vu
H
Kellas
G
Metcalf
K
Strength of discourse context as a determinant of the subordinate bias effect
Q J Exp Psychol
,
1999
, vol.
52
(pg.
813
-
839
)
Mason
RA
Just
MA
Lexical ambiguities in sentence comprehension
Brain Res
,
2007
, vol.
1146
(pg.
115
-
127
)
Mestres-Missé
A
Càmara
E
Rodriguez-Fornells
A
Rotte
M
Münte
TF
Functional neuroanatomy of meaning acquisition from context
J Cogn Neurosci
,
2008
, vol.
20
(pg.
1
-
14
)
Milberg
W
Blumstein
S
Giovanello
KS
Misiurski
C
Summation priming in aphasia: evidence for alterations in semantic integration and activation
Brain Cogn
,
2003
, vol.
51
(pg.
31
-
47
)
Moritz
S
Mersmann
K
Quast
C
Andresen
B
Assoziationsnormen für 68 deutsche Homonyme
Z Exp Psychol
,
2001
, vol.
48
(pg.
226
-
238
)
Moss
HE
Abdallah
S
Fletcher
P
Bright
P
Pilgrim
L
Acres
K
Tyler
LK
Selecting among competing alternatives: selection and retrieval in the left inferior frontal gyrus
Cereb Cortex
,
2005
, vol.
15
(pg.
1723
-
1735
)
Neely
JH
Besner
D
Humphreys
GW
Semantic priming effects in visual word recognition: a selective review of current findings and theories
,
1991
Hillsdale, NJ
Erlbaum
(pg.
264
-
336
)
Noppeney
U
Patterson
K
Tyler
LK
Moss
H
Stamatakis
EA
Bright
P
Mummery
C
Price
CJ
Temporal lobe lesions and semantic impairment: a comparison of herpes simplex virus encephalitis and semantic dementia
Brain
,
2007
, vol.
130
(pg.
1138
-
1147
)
Noppeney
U
Phillips
J
Price
C
The neural areas that control the retrieval and selection of semantics
Neuropsychologia
,
2004
, vol.
42
(pg.
1269
-
1280
)
Obleser
J
Wise
RJS
Dresner
MA
Scott
SK
Functional integration across brain regions improves speech perception under adverse listening conditions
J Neurosci
,
2007
, vol.
27
(pg.
2283
-
2289
)
Onifer
W
Swinney
DA
Accessing lexical ambiguities during sentence comprehension: effects of frequency of meaning and contextual bias
Mem Cogn
,
1981
, vol.
9
(pg.
225
-
236
)
Peleg
O
Eviatar
Z
Hemispheric sensitivities to lexical and contextual information: evidence from lexical ambiguity resolution
Brain Lang
,
2008
, vol.
105
(pg.
71
-
82
)
Peleg
O
Giora
R
Fein
O
Salience and context effects: two are better than one
Metaphor Symbol
,
2001
, vol.
16
(pg.
173
-
192
)
Price
CJ
The anatomy of language: contributions from functional neuroimaging
J Anat
,
2000
, vol.
197
(pg.
335
-
359
)
Pylkkänen
L
Marantz
A
Tracking the time course of word recognition with MEG
Trends Cogn Sci
,
2003
, vol.
7
(pg.
187
-
189
)
Ramnani
N
Owen
AM
Anterior prefrontal cortex: Insights into function from anatomy and neuroimaging
Nature Reviews. Neuroscience
,
2004
, vol.
5
(pg.
184
-
194
)
Raposo
A
Moss
HE
Stamatakis
TA
Tyler
LK
Repetition suppression and semantic enhancement: an investigation of the neural correlates of priming
Neuropsychologia
,
2006
, vol.
44
(pg.
2284
-
2295
)
Rayner
K
Frazier
L
Selection mechanisms in reading lexically ambiguous words
J Exp Psychol Learn Mem Cogn
,
1989
, vol.
15
(pg.
779
-
790
)
Rayner
K
Pacht
JM
Duffy
SA
Effects of prior encounter and global discourse bias on the processing of lexically ambiguous words: evidence from eye fixations
J Mem Lang
,
1994
, vol.
33
(pg.
527
-
544
)
Rissman
J
Eliassen
JC
Blumstein
SE
An event-related fMRI investigation of implicit semantic priming
J Cogn Neurosci
,
2003
, vol.
15
(pg.
1160
-
1175
)
Rodd
JM
Davis
MH
Johnsrude
IS
The neural mechanism of speech comprehension: fMRI studies of semantic ambiguity
Cereb Cortex
,
2005
, vol.
15
(pg.
1261
-
1269
)
Roskies
AL
Fiez
JA
Balota
DA
Raichle
ME
Petersen
SE
Task-dependent modulation of regions in the left inferior frontal cortex during semantic processing
J Cogn Neurosci
,
2001
, vol.
13
(pg.
829
-
843
)
Rossell
SL
Price
CJ
Nobre
AC
The anatomy and time course of semantic priming investigated by fMRI and ERPs
Neuropsychologia
,
2003
, vol.
41
(pg.
550
-
564
)
Ruff
I
Blumstein
SE
Myers
EB
Hutchison
E
Recruitment of anterior and posterior structures in lexical-semantic processing: an fMRI study comparing implicit and explicit tasks
Brain Lang
,
2008
, vol.
105
(pg.
41
-
49
)
Salmelin
R
Clinical neurophysiology of language: the MEG approach
Clin Neurophysiol
,
2007
, vol.
118
(pg.
237
-
254
)
Salmelin
R
Kujala
J
Neural representations of language: activation versus long-range connectivity
Trends Cogn Sci
,
2006
, vol.
10
(pg.
519
-
525
)
Sass
K
Krach
S
Sachs
O
Kircher
T
Lion-tiger-stripes: neural correlates of indirect semantic priming across processing modalities
Neuroimage
,
2008

Available online October 29, 2008, doi: 10.1016/j.neuroimage.2008.10.014
Scheperjans
F
Grefkes
C
Palomero-Gallagher
N
Schleicher
A
Zilles
K
Subdivisions of human parietal area 5 revealed by quantitative receptor autoradiography: a parietal region between motor, somatosensory, and cingulate cortical areas
Neuroimage
,
2005
, vol.
25
(pg.
975
-
992
)
Scheperjans
F
Hermann
K
Eickhoff
SB
Amunts
K
Schleicher
A
Zilles
K
Observer-independent cytoarchitectonic mapping of the human superior parietal cortex
Cereb Cortex
,
2007
, vol.
18
(pg.
846
-
867
)
Scheperjans
F
Palomero-Gallagher
N
Grefkes
C
Schleicher
A
Zilles
K
Transmitter receptors reveal segregation of cortical areas in the human superior parietal cortex: relations to visual and somatosensory regions
Neuroimage
,
2005
, vol.
28
(pg.
362
-
379
)
Scott
SK
Leff
AP
Wise
RJ
Going beyond the information given: a neural system supporting semantic interpretation
Neuroimage
,
2003
, vol.
19
(pg.
870
-
876
)
Sereno
SC
Brewer
CC
O'Donnell
PJ
Context effects in word recognition: evidence for early interactive processing
Psychol Sci
,
2003
, vol.
14
(pg.
328
-
333
)
Sieboerger
FT
Ferstl
EC
von Cramon
DY
Making sense of nonsense: an fMRI study of task induced inference processes during discourse comprehension
Brain Res
,
2007
, vol.
1166
(pg.
77
-
91
)
Simpson
GB
Meaning dominance and semantic context in the processing of lexical ambiguity
J Verbal Learn Verbal Behav
,
1981
, vol.
20
(pg.
120
-
136
)
Simpson
GB
Gernsbacher
MA
Context and the processing of ambiguous words
Handbook of psycholinguistics
,
1994
San Diego, CA
(pg.
359
-
374
)
Slotnick
SD
Moo
LR
Segal
JB
Hart
J
Jr
Distinct prefrontal cortex activity associated with item memory and source memory for visual shapes
Cogn Brain Res
,
2003
, vol.
17
(pg.
75
-
82
)
Small
DM
Gitelman
DR
Gregory
MD
Nobre
AC
Parrish
TB
Mesulam
M-M
The posterior cingulate and medial prefrontal cortex mediate the anticipatory allocation of spatial attention
Neuroimage
,
2003
, vol.
18
(pg.
633
-
641
)
Speer
NK
Zacks
JM
Reynolds
JR
Human brain activity time-locked to narrative event boundaries
Psychol Sci
,
2007
, vol.
18
(pg.
449
-
455
)
Spitzer
M
Braun
U
Hermle
L
Maier
S
Associative semantic network dysfunction in thought-disordered schizophrenic patients: direct evidence from indirect semantic priming
Biol Psychiatry
,
1993
, vol.
34
(pg.
864
-
877
)
Tewes
U
HAWIE-R: Hamburg-Wechsler Intelligenztest für Erwachsene; Handbuch und Testanweisung
,
1991
Bern, Switzerland
Huber-Verlag
Thompson-Schill
SL
Neuroimaging studies of semantic memory: inferring “how” from “where
Neuropsychologia
,
2003
, vol.
41
(pg.
280
-
292
)
Thompson-Schill
SL
D'Esposito
M
Aguirre
GK
Farah
MJ
Role of the left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation
,
1997
, vol.
94
(pg.
14792
-
14797
)
Thompson-Schill
SL
D'Esposito
M
Kan
IP
Effects of repetition and competition on activity in the left prefrontal cortex during word generation
Neuron
,
1999
, vol.
23
(pg.
513
-
522
)
Thompson-Schill
SL
Swick
D
Farah
MJ
D'Esposito
M
Kan
IP
Knight
RT
Verb generation in patients with focal frontal lesions: a neuropsychological test of neuroimaging findings
,
1998
, vol.
95
(pg.
15855
-
15860
)
Vihla
M
Laine
M
Salmelin
R
Cortical dynamics of visual/semantic vs. phonological analysis in picture confrontation
Neuroimage
,
2006
, vol.
33
(pg.
732
-
738
)
Vogt
BA
Finch
DM
Olson
CR
Functional heterogeneity in cingulate cortex: the anterior executive and posterior evaluative regions
Cereb Cortex
,
1992
, vol.
2
(pg.
435
-
443
)
Wagner
Paré-Blagoev
EJ
Clark
J
Poldrack
RA
Recovering meaning: left prefrontal cortex guides controlled semantic retrieval
Neuron
,
2001
, vol.
31
(pg.
329
-
338
)
Wagner
Shannon
BJ
Kahn
I
Buckner
RL
Parietal lobe contributions to episodic memory retrieval
Trends Cogn Sci
,
2005
, vol.
9
(pg.
445
-
451
)
West
R
Jakubek
K
Wymbs
N
Perry
M
Moore
K
Neural correlates of conflict processing
Exp Brain Res
,
2005
, vol.
167
(pg.
38
-
48
)
Xu
J
Kemeny
S
Park
G
Frattali
C
Braun
A
Language in context: emergent features of word, sentence, and narrative comprehension
Neuroimage
,
2005
, vol.
25
(pg.
1002
-
1015
)
Zempleni
M-Z
Renken
R
Hoeks
JCJ
Hoogduin
JM
Stowe
LA
Semantic ambiguity processing in sentence context: evidence from event-related fMRI
Neuroimage
,
2007
, vol.
34
(pg.
1270
-
1279
)