|Home | About | Journals | Submit | Contact Us | Français|
We investigated the role of prefrontal cortex (PFC) in local contextual processing using a combined event-related potentials and lesion approach. Local context was defined as the occurrence of a short predictive series of visual stimuli occurring before delivery of a target event. Targets were preceded by either randomized sequences of standards or by sequences including a three-stimulus predictive sequence signalling the occurrence of a subsequent target event. PFC lesioned patients were impaired in their ability to use local contextual information. The response time for controls revealed a larger benefit for predictable targets than for random targets relative to PFC patients. PFC patients had reduced amplitude of a context-dependent positivity and failed to generate the expected P3b latency shift between predictive and non-predictive targets. These findings show that PFC patients are unable to utilize predictive local context to guide behaviour, providing evidence for a critical role of PFC in local contextual processing.
Contextual processing is essential for the performance of cognitive functions, ensuring that we are able to flexibly adapt our behaviour according to the requirements of particular goals or tasks. Contextual processing enables extraction of relevant environmental information to facilitate the selection of appropriate task-specific responses. For example, we process local contextual information every time we drive our car and see a traffic light turn from green to yellow to red. Utilizing this sequence of events allows us to choose the appropriate response to hit the brakes and to stop our vehicle. Evidence from neuropsychological, event-related potential (ERP) and neuroimaging studies supports a key role of the prefrontal cortex (PFC) in contextual processing (MacDonald et al., 2000; Barch et al., 2001; Huettel et al., 2005; Barcelo and Knight, 2007). Studies of patients with schizophrenia with putative PFC dysfunction also reveal impairments in contextual processing (Barch et al., 2001; MacDonald et al. 2005).
Contextual information is proposed to influence working memory processes supported by the lateral PFC (Cohen and Servan-Schreiber, 1992; Barch et al., 2001). The proposition is that lateral PFC recodes information into context representations (Cohen and Servan-Schreiber, 1992; MacDonald et al., 2000). Information such as task instructions or a cue for the processing of sequential stimuli are maintained in, and manipulated by, the PFC to facilitate appropriate response to salient target stimuli (Cohen and Servan-Schreiber, 1992; MacDonald et al., 2000; Barch et al., 2001; Huettel et al., 2005).
Electrophysiological studies provide a link between contextual processing and the P300 component of the ERP (Squires et al., 1976; Donchin and Coles, 1988; Poulsen et al., 2005; Polich and Criado, 2006; Barcelo and Knight, 2007). The target P300, known as the P3b, is elicited by the classical oddball target detection task and has a posterior–parietal scalp distribution (Squires et al., 1975). P3b latency is a measure of the timing of mental processes reflected by the component, whereas P3b amplitude has been proposed to reflect the intensity of these processes (Kok, 2001). P3b is thought to be a measure of the evaluation of environmental signals including contextual information (Squires et al., 1976; Donchin and Coles, 1988). Short-term local processing of informative stimuli, delivered before the occurrence of target events, has been reported to affect the P3b such that the predictive local context reduces the duration of stimulus evaluation of predicted targets compared with random targets, as seen by a P3b latency shift (Fogelson et al., 2009). This local predictability effect has also been shown to be associated with the generation of a context-dependent positivity (CP), observed only during predictive target detection compared with random target detection. The CP was generated substantially earlier than the conventional P3b (200 ms; Fogelson et al., 2009). These electrophysiological findings support the existence of multiple neural systems in the service of local contextual processing. Studies reporting P3b amplitude suppression in patients with schizophrenia (Alain et al., 1988; Frodl-Bauch et al., 1999) and prefrontal lesions (Barcelo et al., 2000, 2007) provide a link between the P300 family of ERPs to the guided activation model of frontal function (Miller and Cohen, 2001; Barcelo and Knight, 2007). Evidence of the modality-independent effect of local predictive context on behaviour (Fogelson et al., 2009) suggests a further link between contextual processing and PFC.
Although lateral PFC has been implicated in local contextual processing, the critical role of PFC remains unsubstantiated. The aim of the present study was to use a combined ERP and lesion approach to determine whether PFC is a critical brain region for local contextual processing. To this end, we used electrophysiological measures to assess the effects of unilateral prefrontal lesions on the processing of local contextual information. Local context was defined as the occurrence of a short predictive series of visual stimuli before the delivery of a target event. We used the P3b ERP component and the CP to examine the effects of a predictive sequence on electrophysiological measures of local contextual processing in patients who had unilateral lesions in the lateral PFC. We employed a variant of a previously reported paradigm (Fogelson et al., 2009), which showed that predictive local context affects target detection by reducing the duration of stimulus evaluation. This effect was associated with faster reaction times, shortened P3b latencies and generation of a newly described CP seen only during predictive local context-dependent target detection compared with random target detection.
Our hypothesis was that patients with prefrontal lesions would be impaired in behavioural as well as electrophysiological indices of short-term contextual processing. First, we predicted that local contextual information would not facilitate reaction times to salient events in patients. Second, we predicted that the P3b latency shift observed between predicted and random targets in earlier studies in controls (Fogelson et al., 2009) would be abolished in PFC patients, indicating an inability to utilize predictive local context in order to reduce stimulus evaluation time. Finally, we predicted that the early latency CP would be reduced in patients compared with controls, indicating impaired early processes supporting the utilization of the contextual information.
Seven patients (mean age ± SD 60.7 ± 10.9 years, three females) and seven age-matched controls (mean age ± SD 59.3 ± 9.1, three females) participated in the study. Patients had unilateral focal lesions restricted to lateral PFC (Fig. 1). Six of the cases were lesions due to ischaemic infarcts and one due to a haemorrhagic infarct. There were three right and four left lesioned patients. Testing took place at least 4 years after injury. Medical complications, psychiatric disturbance, substance abuse, psychoactive drug treatment or other neurological diseases were criteria for exclusion. All patients had normal or corrected-to-normal visual acuity. Six patients had modest upper motor neuron weakness in the limbs contralateral to their lesions. Patients were matched by seven controls for age, sex and education. Age-matched controls had normal or corrected-to-normal visual acuity and had no history of psychiatric or neurological problems. Subjects were consented prior to being tested and were paid for their participation. The Committee for the Protection of Human Subjects for University California, Berkeley and the Human Subjects Review Committees of the Martinez Veterans Administration Research Service approved the study.
Subjects sat in a sound-attenuated booth 110 cm in front of a 21-in. CRT PC-computer screen. Stimuli were presented to either the left or right visual field 6° from fixation. The stimuli consisted of black triangles on a grey background (Fig. 2). The subject was asked to centrally fixate throughout the recording. Stimuli consisted of 15% targets (downward-facing triangle) and 85% of equal amounts of three types of standards (triangles facing left, upward and right, at 90° increments). In each block, a total of 78 stimuli (12 targets, 22 of each standard type) were presented each for 150 ms and ISI of 1 s. Recording blocks consisted of targets preceded by either randomized sequences of standards or by sequences including a three-standard predictive sequence. The predictive sequence always consisted of the three standards of triangles facing left, up and right, always in that order. Figure 2 illustrates an example of randomized and predicted sequences. Each block consisted six different randomized sequences of standards (3–8 standards long) preceding the target; and six sequences of standards (3–8 standards long) with a predictive sequence preceding the target in each. Each recording session consisted of 14 different blocks, displayed in randomized order, each ~1.6 min long. Blocks were counterbalanced such that there were equal amount of stimuli presented to the right and left visual hemifield across the blocks. Within a single sequence of trials (predictive or random), all stimuli were presented in one hemifield.
Subjects performed a brief training session to ensure that they were able to detect the target accurately. Subjects were then shown the predictive sequence before the recordings began and were told that it would be 100% predictive of a target, but that targets would also appear randomly throughout the block. Subjects were asked to press a button each time a target was presented and to pay attention and look for the predictive sequence. Subjects then performed another brief training session to ensure that they were confident in the detection of the predictive sequence as well as the targets. To respond, the subjects used their best hand (right hand in all the subjects except three patients who used their left hand). Stimulus presentation and response recordings were controlled using E-prime (Psychology Software Tools, Inc., Pittsburgh, USA).
EEG was recorded from a 64-electrode array using the ActiveTwo system (Biosemi, The Netherlands). External electrodes above and below the right eye monitored vertical eye movements and the electrodes placed laterally to the left and right eyes monitored horizontal eye movements. Signals were amplified and digitized at 512 Hz and filtered at 0.16–100 Hz. Post-processing and ERP analysis of the data was performed using Brain Vision Analyser (Brain Products GmbH, Germany). All channels were re-referenced to averaged linked earlobes.
Prior to ERP analysis, ocular movements were defined using ICA and were removed by a linear derivation using Brain Vision Analyser. Epochs containing misses (no button press 150–1150 ms post-stimulus onset) were excluded from further analysis. EEG signals were filtered at 0.1–30 Hz for subsequent analysis. EEG signals were sorted and averaged relative to the stimulus onset, with epochs set from −200 to 1000 ms relative to stimulus onset. EEG epochs with amplitude of >75 μV at any electrode were excluded.
As behavioural and electrophysiological data were comparable for left and right lesioned prefrontal patients, reaction times and electrophysiological data are presented at electrode sites ipsilateral and contralateral to the lesion and for stimuli delivered in the visual field ipsilateral or contralateral to the lesion. Ipsilesional electrode sites are denoted with i and contralesional electrode sites are denoted with c. Thus, Fi, Ci, Pi and POi refer to the averaged ERP data from electrode sites F3 or F4, C3 or C4, P3 or P4 and PO7 or PO8, for left and right prefrontal lesions, respectively. While, Fc, Cc, Pc and POc refer to the averaged ERP data from electrode F3 or F4, C3 or C4, P3 or P4 and PO7 or PO8, for right and left prefrontal lesions, respectively. Ipsilesional ERP data were compared with left hemisphere data from controls, and vice versa. No significant differences were observed between right and left lesioned patients. However, laterality effects cannot be ruled out due to the power limitations of the present study.
P3b was determined as the most positive point in the latency range of 300–800 ms. Detected peaks were then checked for their topographical map (using Brain Vision Analyser) and confirmed as a P3b component by determining a posterior–parietal scalp distribution. In order to restrict the number of comparisons and to determine if any significant lateralization effects existed, an omnibus analysis of variance (ANOVA) was first performed using the P3b peak amplitude at electrode sites Fc, Fi, Fz, Cc, Ci, Cz, Pc, Pi, Pz in the patients and at F3, F4, Fz, C3, C4, Cz, P3, P4, Pz in the controls and from both visual fields of stimuli presentation (ipsilesional and contralesional in the patients or left and right in controls) for four different conditions (predicted and random targets, random standards and the last most-informative standard comprising the predicting sequence) across both groups (patients and controls). No differences were observed between the two visual fields of presentation, or between ipsilesional or left and contralesional or right hemisphere electrode sites. Independence of visual field of presentation has been shown in other studies (Duarte et al., 2005; Barcelo and Knight, 2007). Thus, all P3b ERP data were collapsed across visual fields for subsequent analysis. In addition, we concentrated on midline electrode sites (AFz, Fz, FCz, Cz, CPz and Pz) to explore anterior versus posterior topographical differences of P3b for the different target conditions.
Peak P3b amplitude (measured in microvolt) and latencies (measured in milliseconds) at AFz, Fz, FCz, Cz, CPz and Pz were evaluated for four conditions: targets after predictive sequences (predicted), targets after non-predictive random sequences (random), random preceding standards (standards excluding those comprising of the predicting sequence) and the last most informative standard of the predicting sequence (n–1) for both groups.
A difference wave subtracting random targets from predicted targets was derived to extract the CP that was observed in the predicted target compared with the random target condition. Peak amplitude was evaluated for this difference wave by determining the most positive point in the latency range of 200–400 ms. Like the P3b, no field effects were observed and data were collapsed across visual fields for subsequent analysis. The peak CP was evaluated at AFz, Fz, FCz, Cz, CPz and Pz.
To assess the early perceptual processes between the two target conditions, peak N1 amplitudes (measured in microvolt) were determined at PO7 and PO8 in controls and at POi and POc in the patients, for both predicted and random targets presented to the right or left visual field in the controls and to the contralesional and ipsilesional visual fields in the patients. N1 was determined as the most negative peak in the latency range of 50–200 ms.
ANOVA was performed with the Greenhouse-Geisser correction, followed by post hoc parametric paired t-tests, Sidak corrected for multiple comparisons unless otherwise stated. Mean values ± standard error of the mean (SEM) are used throughout the text.
Mean accuracy was 98 ± 0.7% and 99 ± 0.3% (P = NS) for the PFC patients and age-matched controls, respectively. To compare the reaction times (RTs) for the targets between the groups, we performed an ANOVA with group (PFC patients, controls) as the between-subject factor and condition (predicted, random targets) as the repeated measure factor. There was a main effect for condition [F(1,12) = 28.65, P < 0.0001] and a significant condition × group interaction [F(1,12) = 5.59, P = 0.036]. Post hoc t-tests showed that in controls, RTs for predicted targets (mean RT = 370 ± 15 ms) were shorter than those for random targets [mean RT = 510 ± 24 ms, t(6) = 4.92, P = 0.003]. In the PFC patients, RTs for predicted targets (mean RT = 470 ± 35 ms) were not significantly different than random targets [mean RT = 524 ± 23 ms, t(6) = 2.41, P = NS].
Scalp distributions of the grand-averaged ERPs across the seven subjects in the control group and seven subjects in the patient group at electrodes CPz and Pz elicited by predicted and random targets, standards and n–1, and the last most informative stimulus of the predicting sequence, are shown for the patients and controls in Fig. 3A and B.
To compare peak P3b amplitudes, we performed an ANOVA with group (PFC patients, controls) as the between-subject factor and with condition (predicted, random targets, n–1 and standards) and electrode (AFz, Fz, FCz, Cz, CPz and Pz) as the repeated measures factors. There was a main effect for condition [F(3,36) = 25.14, P < 0.0001] and electrode [F(5,60) = 27.58, P < 0.0001]. There was no effect of group nor were there group interactions with any of the factors in the comparison of peak P3b amplitude.
Both groups showed maximal and minimal P3b amplitudes at electrode sites CPz and AFz, respectively. Post hoc tests, corrected for multiple comparisons, showed that the peak P3b amplitude in the control group was larger for predicted targets, random targets and n–1 condition compared with standards (P ≤ 0.026). In the patients, peak P3b amplitude was larger for predicted and random targets when compared with the n–1 and standards condition (P < 0.0005). Thus, unlike controls, PFC patients failed to generate a robust P3b to the n–1 condition (Fig. 3A and B). P3b amplitude was comparable for random and predicted target stimuli in both controls and PFC patients (Fig. 3A and B).
Since peak P3b amplitudes were largest at CPz, we used this electrode in the comparison of peak P3b latencies and performed an ANOVA with group (PFC patients, controls) as the between-subject factor and with condition (predicted, random targets) as the repeated measures factor. There was a main effect for the condition [F(1,12) = 9.53, P = 0.009] and a significant condition × group interaction [F(1,12) = 6.43, P = 0.026]. An independent t-test showed a significant P3b latency shift between predicted and random targets in age-matched controls compared with PFC patients [t(12) = 2.5, P = 0. 026]. In controls, the peak P3b latency was shorter for predicted targets (mean P3b latency = 538 ± 49 ms) when compared with the peak P3b latency for random targets [mean P3b latency = 652 ± 47 ms, t(6) = 3.1, P = 0.022]. However, in the patients, peak P3b latency did not differ between predicted (mean P3b latency = 539 ± 43 ms) and random targets (mean P3b latency = 551 ± 32 ms; P = NS, Fig. 3A and B). There were no significant differences in peak P3b latency between patients and controls for both target conditions.
The CP is generated to targets preceded by a predictive sequence. The CP is derived by calculating the difference wave between predicted and random targets. We compared the peak CP amplitude, using an ANOVA with group (PFC patients, controls) as the between-subject factor and electrode (AFz, Fz, FCz, Cz, CPz and Pz) as the repeated measures factor. There was a main effect for electrode [F(5,60) = 16.8, P < 0.0001] and a significant electrode × group interaction [F(5,60) = 8.21, P = 0.001]. Peak CP amplitudes were maximal at CPz and Pz in the patients and controls, respectively, and minimal at AFz in both groups. Thus, the CP has a posterior–parietal scalp distribution similar to that of the P3b (Fig. 3C and and4B).4B). Attenuation of peak CP amplitudes was observed at CPz and Pz in patients compared with controls. Independent t-tests revealed maximal CP amplitudes at Pz, with a reduced peak CP amplitude in the patients (mean CP amplitude = 5.01 ± 1.45 μV) compared with the controls [mean CP amplitude = 10.57 ± 0.83 μV, t(12) = 3.33, P = 0.006], as well as an attenuation of peak CP amplitude at CPz in the patients (mean CP amplitude = 5.86 ± 1.72 μV) compared with the controls [mean CP amplitude = 10.52 ± .94 μV, t(12) = 2.38, P = 0.035] (as illustrated in Fig. 4A). CP amplitudes at Pz and CPz were smaller for stimuli delivered to the visual field contralateral to the lesion in patients (mean CP amplitude = 6.43 ± 1.28 μV and 6.74 ± 1.55 μV for Pz and CPz, respectively) compared with CP amplitudes for stimuli delivered to controls [mean CP amplitude = 11.57 ± 1.13 μV, t(12) = 3.02, P = 0.011 and 11.41 ± 0.98 μV, t(12) = 2.55, P = 0.025; Fig. 5]. CP amplitudes for ipsilesional stimuli (mean CP amplitude at CPz = 8.36 ± 2.33 μV) did not differ from controls (mean CP amplitude = 11.47 ± 1.58 μV, P = NS).
Differences between the CP and P3b components were evaluated by comparing their peak latencies. t-Tests comparing CP peak latency versus P3b peak latency for targets at Pz in controls showed CP peak latency to be substantially shorter (mean CP latency = 342 ± 14 ms) than P3b peak latency for both predicted [mean P3b latency = 541 ± 50 ms, t(6) = 3.69, P = 0.01] and random targets [mean P3b latency = 649 ± 47 ms, t(6) = 6.78, P = 0.001].
We utilized an ANOVA with group (patients, controls) as the between-subject factor and electrode (POi/PO7 and POc/PO8), visual hemifield (ipsilesional/left, contralesional/right) and condition (predicted, random targets) as repeated measures factors, to compare the peak N1 amplitude between predicted and random targets. There was no significant overall effect between the groups in the comparison of peak N1 amplitude. There was a main effect for condition [F(1,12) = 13.23, P = 0.003]. However, post hoc t-tests did not reveal significant differences between predicted and random targets when comparing within one electrode (POi/PO7 or POc/PO8) and within one visual field (ipsilesional/left and contralesional/right) in both groups. There was a significant interaction between extrastriate electrode locations and visual field of stimuli presentation [F(1,12) = 35.65, P < 0.0001] across both groups, demonstrating that N1 ERPs were enhanced to contralaterally presented stimuli.
The current study provides behavioural and electrophysiological evidence that patients with lateral prefrontal lesions are impaired in their ability to process local contextual information. These results provide key evidence for the proposal that lateral PFC is critical for contextual processing. This was demonstrated by three main electrophysiological findings. PFC patients demonstrated a prominent reduction of the peak of an early latency context-dependent positivity compared with controls. Further, patients did not show the P3b latency shift between predicted and non-predictive targets that was observed in controls, indicating that there was no differential processing of predicted versus random targets in the patients, while controls utilized predictive local context to reduce the duration of stimulus evaluation (Fogelson et al., 2009). Unlike controls, PFC patients failed to generate a robust P3b to the last most informative stimulus (n–1) of the predictive sequence. Finally, local contextual information did not facilitate behavioural performance.
In the controls, we replicated findings of a previous study (Fogelson et al., 2009) showing that P3b amplitude increased with task-informative stimuli. Importantly, a significant P3b is generated by the last and most-informative stimulus of the predicting sequence, and P3b amplitude then reaches a maximum for target events. These findings support the widely held view that P3b amplitude increases as a function of task relevance (Sawaki and Katayama, 2006; Fogelson et al., 2009) and that n–1 became a secondary target for the subjects and thus an indicator for successful local contextual processing. PFC patients had comparable P3b amplitudes for the predicted and random targets. However, unlike the controls, n–1 did not induce a robust P3b, suggesting that PFC patients were unable to attend, maintain and utilize the contextual information provided by the predictive sequence. However, as noted, there were no significant differences in P3b amplitude between predicted and random targets in both groups, suggesting that in this easy discrimination task, what determined the magnitude of the P3b amplitude was the task relevance of the stimulus (replicating Fogelson et al., 2009).
In controls, P3b latency was shorter for sequence-predicted targets than for targets after non-predictive sequences (Fogelson et al., 2009) indicating that predictive local context affects the target detection by reducing the duration of stimulus evaluation (Kutas et al., 1977; Duncan-Johnson, 1981; McCarthy and Donchin, 1981; Duncan-Johnson and Donchin, 1982; Hillyard and Kutas, 1983; Fogelson et al., 2009). In patients with prefrontal lesions, P3b latency did not differ between predicted and random targets, suggesting that patients processed predicted and random targets similarly, while control subjects utilized local predictive contextual information to speed cognitive processing. These results were associated with parallel behavioural findings, which revealed that in controls, there is a shortening in RTs for predictables compared with random targets, relative to PFC patients. Similar behavioural results of a failure to utilize novel information after PFC damage have also been reported (Barcelo and Knight, 2007). Another possible interpretation for the lack of difference between random and predicted targets in the patients may be attributed to a lack of inhibition to unexpected events. However, this seems unlikely, since there were no significant differences in P3b latencies between the two groups. The current findings suggest that the impairment in the patients and the facilitation of stimulus evaluation in the controls is cognitive rather than perceptual, since targets were identical in their physical features and there were no significant N1 amplitude differences between predicted and random targets (Hillyard and Kutas, 1983) in both the groups.
The CP was observed only for predicted targets and CP peak was attenuated in patients with PFC lesions compared with controls. CP amplitude reduction was more pronounced when the stimulus was delivered to the lesioned compared to when it was delivered to the intact hemisphere in the patients. The CP has a similar distribution to the P3b. However, it has a shorter peak latency than P3b, being generated ~200 ms earlier than the P3b. This suggests that the CP reflects a distinct neural process generated during local contextual processing. We have suggested that the CP reflects an early template match process related to early target detection processes or preparatory attention of working memory (Squires et al., 1973; Chao et al., 1995; Fogelson et al., 2009). PFC dysfunction has been shown to be associated with working memory deficits (Barch et al., 2001; Müller et al., 2002). Thus, our findings suggest that contextual effects on early target detection are impaired in patients with lateral PFC damage and that this may be related to working memory deficits that support early latency target detection processes. Working memory deficits may be manifested in the inability of PFC patients to maintain contextual information provided by the predictive sequence, or due to the inability to direct attention to the predictive sequence (Rainer et al., 1998), as supported by the failure of PFC patients to generate a robust P3b to n–1 in our study.
Taken together, the findings in the absence of a P3b latency shift, the attenuation of the CP and the lack of P3b to the final stimulus in the predictive sequence in patients with lateral PFC lesions, provide evidence for impaired contextual processing in the patients. This is in line with unilateral prefrontal lesion data from animals (Morgan and LeDoux, 1999) and patients (Barch et al., 2001; Duarte et al., 2005; MacDonald et al., 2005; Barcelo and Knight, 2007) suggesting an association between PFC dysfunction and contextual processing deficits. Our findings inform neural modelling and neuroimaging data implicating the PFC as a critical region for support of contextual processing (Cohen and Servan-Schreiber, 1992; Huettel et al., 2005). These investigators suggest that the lateral PFC is responsible for maintaining, representing and manipulating context information in the service of selecting an appropriate task-specific response (Cohen and Servan-Schreiber, 1992; MacDonald et al., 2000; Miller and Cohen, 2001).
It has been suggested that the PFC is important for extracting regularities such as goals, task rules or context representations (Miller and Cohen, 2001). These context representations, which are thought to be multi-modal in nature (Miller, 1999; Miller and Cohen, 2001; Fogelson et al., 2009), can bias both motor and sensory processing to allow for the selection and execution of appropriate actions (Miller, 1999). In the present study, we have shown that patients with lateral PFC lesions are unable to use the rule to extract contextual information in order to predict those targets which were 100% predictive and that this was manifested in the inability to bias motor output in order to facilitate behaviour.
In conclusion, the current study provides evidence of local contextual processing impairments in patients with lateral PFC lesions, by demonstrating alteration in the behavioural and neural correlates of local contextual processing. This evidence demonstrates the critical role of the PFC in contextual processing.
National Institutes of Health (NINDS Grant NS21135 and PO40813).
We would like to thank Clay Clayworth for lesion reconstructions, Kilian Koepsell for his contributions to the design of the paradigm and Jeffrey Lewis for assistance with programming of the task.