Sequential P3 effects in a Posner’s spatial cueing paradigm: Trial-by-trial learning of the predictive value of the cue

Research paper Acta Neurobiol Exp 2009, 69: 155–167 © 2009 by Polish Neuroscience Society - PTBUN, Nencki Institute of Experimental Biology INTRODUCTION In daily life, humans continuously monitor their actions to ensure that they are appropriate for the environment. In a broader perspective, the “perceptionaction cycle” concept was introduced by Fuster (2003, 2004) to highlight the continuous interplay and evaluation of outcomes between perceptual and executive networks. In this neurocognitive model, the sensory and executive networks interact continuously at different levels of the processing hierarchy. One major consequence of such interaction would be the ability to monitor appropriateness between perception and action. Recently, the feedback that the neural system may generate as a result of a trial outcome has been demonstrated through the Error-Related Negativity component (Gehring et al. 1993, Falkenstein et al. 1995, Holroyd and Coles 2002). Another hypothesis linking frontomedial negativities to the evaluation of outcome has been proposed by Vidal and coauthors (2003). These authors sought to explain the frontomedial negativities that may occur after responses to correct trials (Vidal et al. 2003, Burle et al. 2005); they proposed that post-response negativities could be related to the response evaluation process. Therefore, there are clear neurophysiological signs of trial assessment as a function of trial outcome. The perception of a target stimulus is often preceded by a cue that creates expectations about the features and relevance of the target, and this leads to a more complex view of the perception–action cycle: there is a continuous expectancy bias for certain stimuli and actions, converting the former perception-action cycle Sequential P3 effects in a Posner’s spatial cueing paradigm: Trial-by-trial learning of the predictive value of the cue Carlos M. Gómez*, Angélica Flores, Marcia R. Digiacomo, and Manuel Vázquez-Marrufo Laboratory of Human Psychobiology, Department of Experimental Psychology, University of Seville, Spain, *Email: [email protected] The neurocognitive consequences of correct or incorrect spatial prediction in a sequential S1–S2 paradigm were assessed. Sequential dependence on previous trial outcome (valid or invalid) was assessed by late Event-Related Potentials (ERPs) and behavioural responses. Two different experiments were performed, situating the target in the vertical (Experiment 1) or in the horizontal (Experiment 2) meridian. RTs and late positivities (P3a and P3b) were recorded. ERPs showed that posterior positivity (probably a P3b) was greater in invalid–valid trials than in valid–valid trials but lower than in valid– invalid trials. However, at the frontal electrodes, late positivity (probably a P3a) only appeared in valid–invalid trials, indicating that invalid trials are analyzed as novel-like stimuli. The P3b results suggest trial-by-trial learning of the predictive value of the cue, which needs to be updated as indicated by the pattern of P3b amplitudes: valid–invalid > invalid–valid > valid–valid. Key words: sequential effects, predictive value, trial-by-trial learning, P3a, P3b Correspondence should be addressed to C.M. Gómez, Email: [email protected] Received 08 July 2008, accepted 18 April 2009 156 C.M. Gómez et al. into a preparation-perception-action cycle. A neural signature indicating adequacy of preparation, perception and action might help to validate the hypothesis of a complete preparation–perception–action cycle. One type of stimulus sequence that seems particularly well suited to testing congruency between the expected and the current stimulus is the central cue Posner paradigm. In this paradigm, the central cue may validly (VC trials) or invalidly (IC trials) indicate the spatial position of the upcoming target. If the cue is a valid predictor of the target, there is a benefit in the RT with respect to neutral cues. However, if the target is incorrectly cued, a cost occurs in the RT with respect to neutral cues (Posner 1980). Part of this effect is due to the preparation of the incorrect response for invalidly cued target stimuli (Eimer 1993, Gómez et al. 2004). From neurophysiological recordings, the preparation phase of the proposed cycle is explained by the Contingent Negative Variation (CNV). During the late period of the CNV in central cue Posner paradigms, it is possible to observe task-specific preparatory activation of the motor and sensory areas which are potentially needed to complete the task upon the information conveyed by the S1 (Gómez et al. 2003, 2004). Activation of fronto-medial and fronto-parietal areas can also be observed, probably sustaining endogenous attentional effort during the CNV period (Hopfinger et al. 2000, Brunia and Van Boxtel 2001, Gómez et al. 2003, 2004, 2007, Fan et al. 2007). A consequence of the task-specific preparation indicated by the CNV is that the prepared neural set may be correct or incorrect, depending on whether S1 represents the characteristics of S2 validly or invalidly. The spatial cueing effects on the early ERP components reflecting attentional sensory processing have previously been evaluated by analyzing the modulation of early ERPs to target stimuli that are preceded by valid and invalid cues (Mangun and Hillyard 1991, Eimer 1993, Anllo-Vento 1995, Perchet and García- Larrea 2000, Perchet et al. 2001). The posterior positive components under IC conditions are also larger than those under VC conditions at the time latency of the P3 component (Mangun and Hillyard 1991, Eimer 1993). Recently (Gómez et al. 2008a,b, Digiacomo et al. 2008), the P3b effects were replicated and an increased P3a was also found when IC and VC targets were compared. The latter results suggest that invalidly cued targets are processed in a similar manner to low frequency targets in oddball paradigms, as indicated by the greater P3a and P3b in IC trials. The P3a component is generated as a brain response to stimuli that are novel in comparison to more frequent stimulation. In fact, completely novel stimuli generate a higher-amplitude P3a component than deviant but often-repeated stimuli (Escera et al. 1998, Friedman et al. 2001, Wronka et al. 2008). The larger P3a (Friedman et al. 2001, Dien et al. 2003) indicates that the invalidly cued target is processed as a novel stimulus. In contrast, the P3b is a late positive component with a parietal distribution and typically appears in oddball paradigms. It is well known that the amplitude of P3 is inversely related to the probability of stimulus appearance (Duncan-Johnson and Donchin 1977). According to Donchin and Coles (1988), the higher P3b in invalid trials represents the context-updating operation and subsequent memory storage (Polich 2007). An alternative explanation of the cognitive meaning of P3 was suggested by Verleger and others (2005), who proposed that P3 is related to the neural linkage between stimulus perception and the response to that stimulus. A recent proposal relates the P3b component to the neuroinhibition needed to focus attention on the relevant task, facilitating the interference-free action of memory systems (Polich 2007). For the central cue Posner´s paradigm, the greater P3 in IC than VC trials represents assessment of the adequacy of sensory-motor preparation, sensory perception and action. It has been suggested that the most important component of this assessment is revision of the S1–S2 (cue–target) contingency value (Gómez et al. 2008a). In this approach, it is important to separate anterior and early P3 components of ERPs (P3a) from posterior and late ones (P3b), since source localization studies and scalp current source analysis have allowed early anterior and late posterior sources to be separated under IC and VC conditions (Gómez et al. 2008a,b). An important issue that has scarcely been studied in central cue Posner paradigms is how correct or incorrect prediction in a given trial can induce changes in the processing of the next trial, i.e. sequential effects. A recent behavioral report on the central cue Posner paradigm addresses this point (Jongen and Smulders 2007). These authors found an inter-trial validity effect: the benefit in RTs when compared to neutral cues is higher if a valid trial is preceded by a valid trial than if it is preceded by an invalid one. On the other hand, the cost of an invalid trial is greater if it is preceded by a valid trial than by an invalid one. To our Trial-by-trial learning and P3 157 knowledge, there have been no ERP studies analyzing the sequential effects of inter-trial validity using Event Related Potentials (ERPs). ERP studies could elucidate the timing and neurocognitive effects of inter-trial validity in cueing paradigms. Other effects related to neutral and catch trials, to the effects of alternate or repeated responses and to the inhibition of return are also reported by Jongen and Smulders (2007). The aim of the present report is to analyze the sequential effects of valid S1–S2 trials preceded by other S1–S2s. We expect the outcome of the current trial to affect the behavior and ERPs of the next one. To achieve this objective, two experiments were conducted using the central cue Posner paradigm. One experiment presented central cues orientated vertically and targets situated in the vertical meridian, while the other presented central cues orientated horizontally and targets situated in the horizontal meridian (see Figures 1 and 2). Behavioral activity and EEG were recorded simultaneously. More specifically, the sequential effects on the behavioral responses and the P3a and P3b components were analyzed. Two different types of effects on the processing of a target are expected: (i) effects due to the implicit meanings of spatial cues and to the global predictive values of cues during the experiment (the so-called validity–invalidity effect); and (ii) local effects due to the spatial validity or invalidity of the previous trial. The late positive components allow us to analyze how novel the stimuli are considered (via P3a) and how much the internal model of the current predictive value associated with the cue presented (S1–S2 contingency value) needs to be revised (via P3b). Consistency of the results between the two experiments, in which the shapes and positions of targets (vertical vs. horizontal) and the S1–S2 interval (random vs. fixed duration) differed, would indicate robustness. As we are interested only in robust outcomes, we will discuss the results of both experiments together. METHODS Experiment 1 Subjects Fourteen subjects (9 female and 5 male, 10 righthanded) between 21 and 36 years old (mean 24.5) took part in the experiment. The experiments were conducted with the informed and written consent of each subject following the rules of the Helsinki Convention. The Ethics Committee of the University of Seville approved the study. Stimuli and behavioral paradigm The stimulus presentation was computer-controlled (EEVOKE, ANT). Participants were seated 60 cm in front of a computer screen. They were instructed to fix their eyes on a white square in the centre of the screen, which remained visible throughout the experiment in order to sustain central fixation. The complete trial period included a central directional cue that was on for 200 ms, then an attentive waiting period lasting randomly between 1 800 and 2 000 ms (Fig. 1). Finally, a peripheral target appeared, subtending a visual angle of 0.91º and situated 8.3º eccentrically in the vertical meridian. Since the central directional cue could indicate the direction of appearance of the target correctly or incorrectly, two different conditions arose: validly cued targets (VC) (82.13% of trials) and invalidly cued targets (IC) (17.87% of trials). In the present report, we will focus on the behavioral and ERP effects of valid and invalid trials that were preceded by validly or invalidly cued trials. We will Fig. 1. Paradigm for Experiment 1. The different types of sequence trials considered in the experiment are shown. The temporal sequence of stimulus presentation appears in the lower part of the figure. Notice that the RTs and ERPs were obtained from the S2 stimulus in the second trial. This corresponds to the stimulus right side of the figure in each stimulus sequence. 158 C.M. Gómez et al. therefore consider four types: S1–S2 valid trials preceded by a valid one (VV) (159 trials, 67.66% of the total); valid trials preceded by an invalid one (IV) (34 trials, 14.47% of the total); invalid trials preceded by a valid one (VI) (34 trials, 14.47% of the total); and invalid trials preceded by an invalid one (II) (8 trials, 3.40% of the total). The most relevant comparison in the present report is between IV and VV: there were too few II trials for reliable results, and comparisons between invalidly-cued trials and validly-cued trials have already been described (Mangun and Hillyard 1991, Eimer 1993, Perchet and García-Larrea 2000, Perchet et al. 2001, Gómez et al. 2008a,b, Digiacomo et al. 2008). The participants used their right hands to respond to the targets by pressing a joystick button. They used the index finger to respond to targets presented in the upper visual hemifield and the thumb to respond to targets presented in the lower one. The inter-trial intervals were randomly selected between 2 800 and 3 000 ms. Subjects were presented with a total of 235 trials. There were no training trials. Behavioral statistical analysis The RTs were measured for VV, IV, VI and II trials and analyzed by a one-factor ANOVA with four levels. The Bonferroni test was used for planned a priori comparisons. As an indicator of the subjects’ general performances, the errors in individual trials were computed regardless of the type of the previous trial. Responses to S1, omissions to S2 and response errors to S2 were computed independently for validly and invalidly cued trials. The total numbers of errors in the two trial types were compared statistically by repeated measures ANOVA EEG recording, processing and analysis The EEG was recorded from 64 scalp sites in an extended version of the International 10–20 system, using tin electrodes mounted in an electrode cap (electrocap). All the electrodes were referred to the left mastoid. Impedance was maintained below 5 000 Ohms. Data were recorded in DC and no filtering was applied to them. The amplification gain was 20 000 (ANT amplifiers). The data were acquired at a sampling rate of 512 Hz, using a commercial AD acquisition and analysis board (ANT). Recordings were averaged off-line using an artifact-rejection protocol based on voltage amplitude. All the epochs for which the EEG exceeded 100 microvolts in any channel were automatically discarded for ERP analysis; 56.51% of VV trials, 56.72% of IV trials, 64.29% of II trials and 59.87% of VI trials were accepted for analysis. Rejections resulted not only from blinks and eye movements, but also from behavioral errors in the sequences of the two trials, since correct responses were needed in both the actual and the preceding trial. Moreover, not only should the double trial be behaviorally correct, it should not present artifacts in either trial sequence. Also, the first trial in each block (the experiment had five blocks) had to be rejected because there was no preceding trial. ERPs were obtained for each subject by averaging the EEG, using the switching-on of the target as a trigger. The baseline was the interval 200–0 ms before target stimulus. The algebraically linked mastoids were computed off-line and used as reference for analysis purposes. Eye movements and blinks were monitored using the electrodes installed in the 64-channel cap, after the sensitivity of the AF7– AF8 in monitoring horizontal eye movements and of the prefrontal electrodes (Fp1, Fpz and Fp2) in monitoring vertical eye movements and blinks had been checked. Statistical analysis of ERPs Repeated measures ANOVA was performed on the voltage data from selected electrodes. The mean voltage in selected time windows was analyzed independently for P3a and P3b under VV, IV and VI conditions. The II condition was not included in the analysis because there were so few II trials after the artifact rejection protocol had been implemented. The P values were calculated using the Greenhouse-Geisser correction. The Bonferroni method was used as post-hoc test when necessary. Early positivity (P3a) A two factor repeated measures ANOVA was performed on the voltage data with selected midline electrodes (Fz, Fcz, Cz). The mean voltage in the P3a time window (280–340 ms) was computed for each of the three conditions previously described. The factors considered were the type of trial (3 levels: VV, IV and VI) and the electrodes. Trial-by-trial learning and P3 159 Late positivity (P3b) A three factor repeated measures ANOVA was performed on the voltage data with selected electrodes (O1, O2, P3, P4, P5 and P6). The mean voltage in the P3b time window (300–400 ms) was analyzed independently for each of the three conditions considered. The factors considered were the type of trial (3 levels: VV, IV and VI), the hemisphere (left or right) and the electrodes. Experiment 2 The main methodological differences between experiments 1 and 2 were the number of subjects, the location of the stimuli, the response hand, the S1 –S2 period and the number of electrodes. In Experiment 2, the stimuli were in the horizontal meridian and the time between S1 and S2 was fixed. Moreover, responses were produced with both hands. Subjects Sixteen subjects (9 female and 7 male, all 16 righthanded) between 19 and 23 years old (mean 21) took part in the experiment. All experiments were conducted with the informed and written consent of each subject following the rules of the Helsinki Convention. The Ethics Committee of the University of Seville approved the experiment. Stimuli and behavioral paradigm The stimulus presentation was computer-con- trolled (EEVOKE, ANT). Participants were seated 50 cm in front of a computer screen. They were instructed to fix their eyes on a white square in the centre of the screen, which remained visible throughout the experiment in order to sustain central fixation. The complete trial period included a central directional cue that was on for 300 ms, then an attentive waiting period lasting 1 360 ms (Fig. 2). Finally, a peripheral target appeared, subtending a visual angle of 4.56º and situated 3.66º eccentrically in the horizontal meridian. The targets were cartoon figures that were constant in each block of trials. There were five blocks of trials. Five different figures were used in the whole experiment. Since the central directional cue could indicate the direction of appearance of the target correctly or incorrectly, two possibilities arose: validly cued targets (VC) (82.13% of trials) and invalidly cued targets (IC) (17.87% of trials). The numbers and percentages of VV, IV, VI and II trials were identical to those in Experiment 1. The subjects used the right index finger to respond to targets on the right of the screen by pressing a joystick button, and the left index finger to respond to targets on the left. The inter-trial intervals were 1 860 ms. Subjects were presented with a total of 235 trials. There were ten training trials. Behavioral statistical analysis was identical to that used in Experiment 1. EEG recording, processing and analysis The EEG was recorded from 20 scalp sites in an extended version of the International 10–20 system, using tin electrodes mounted in an electrode cap (easy cap). The EEG processing was identical to that in Experiment 1; 62.54% VV, 58.82% IV, 57.81% II and 59.38% VI trials were accepted for ERP analysis. ERPs were obtained for each subject by averaging the EEG, using the switching-on of the target as trigger. The II condition was not included in the statistical analysis because it involved very few trials. Statistical analysis of ERPs was identical to that in Experiment 1. Fig. 2. Paradigm for Experiment 2. The different types of sequence trials considered in the experiment are shown. The temporal sequence of stimulus presentation appears in the lower part of the figure. Notice that the RTs and ERPs were obtained from the S2 stimulus in the second trial. This corresponds to the stimulus on the right side of the figure in each stimulus sequence. 160 C.M. Gómez et al. RESULTS Experiment 1 Behavioral results The RTs in the different sequences of trials appear in Table I. ANOVA showed an effect of the type of trial sequence (F3,39=17.828, P<0.001). The fastest condition was VV, followed by IV, II and finally VI. Bonferroni analysis showed that the attentional cueing manipulation was effective, as indicated by the statistically significant differences between conditions II and VV (P<0.015), II and IV (P<0.033), and VI and both VV (P<0.001) and IV (P<0.001). Table II shows the percentages of the different types of error in the valid and invalid trials. Comparisons showed a significant difference between the percentages of response errors in valid and invalid trials; there were more errors in the invalid trials (P<0.039). No other comparisons (responses to S1 and omissions to S2) were statistically significant. Event-Related Potentials P3a Figure 3 shows higher positivity for the VI condition than the VV and IV conditions. ANOVA showed an effect of the type of trial sequence (F2,26=4.81, P=0.016) and an interaction between the type of trial sequence and the electrodes (F2.26=4,21, P=0.025). Post-hoc analysis showed effects at all the electrodes for the comparisons IV vs. VI (Table III). Comparisons between VV and VI yielded statistically significant differences at the electrodes Fcz and Cz. There were no statistically significant differences between VV and IV at any of the electrodes considered. P3b ANOVA showed that the interaction “type of trial sequence × hemisphere × electrodes” was statistically significant (F10,130=2.75, P=0.004). Post-hoc comparisons, Table I Reaction times of the four experimental stimulus sequences analyzed in Experiment 1 Condition Reaction Times SD Valid–Valid (VV) 372.89 106.67 Invalid–Valid (IV) 379.69 98.04 Invalid–Invalid (II) 413.48 92.95 Valid–Invalid (VI) 430.03 116.78 Table II Percentage of errors in the stimulus sequence of Experiment 1 Condition Percentage of no response SD Percentage/S1 SD Percentage/S2 SD Error Invalid 0.33 0.84 0.17 0.62 1.99 2.56 Valid 0.62 1.15 0.44 0.48 0.29 0.56 Omissions (percentage of no response), false alarms to the S1 of the second trial of the sequence (i.e. S1 before target, indicated as percentage/S1), and false alarms to the target stimuli (percentage/S2) are shown. Table III Post-hoc results for the component P3a in the Experiment 1 Fz Fcz Cz VV vs. VI 0.001 0.006 IV vs. VI 0.008 <0.001 <0.001 VV vs. IV In all cases the Bonferroni correction was applied. The cells with no values were not statistically significant Trial-by-trial learning and P3 161 using the very conservative Bonferroni test, showed statistically significant differences between the IV and VV conditions at the electrodes PO4, PO6 and P8 (Table IV). Figures 3 and 4 show a higher P3b in the IV condition than in VV at the posterior right side of the scalp. Posthoc comparisons between VI and VV or IV were statistically significant at most of the electrodes considered. Experiment 2 Behavioral results The RTs in the different trial sequences appear in Table V. ANOVA showed an effect of the type of trial sequence (F3,45=9.081, P=0.001). The fastest condition was VV, followed by IV, II and finally VI. Bonferroni analysis showed that the RTs in the condition VV were significantly faster than in IV (P< 0.0 46 ) or V I (P<0.001), and that the RTs in the IV condition were significantly faster than in the VI condition (P<0.01). II did not differ from any other condition, probably because there were so few cases of II. The preceding results indicate that the attentional cueing manipulation was effective, but they also show that the IV condition is intermediate in status between the VV and VI conditions. Table VI shows the percentages of different types of error in the valid and invalid trials. Comparisons showed a statistically significant difference between these percentages (Table VI): there were significantly more errors in the invalid trials than in the valid ones Table IV Post-hoc results for the component P3b in the Experiment 1 P3 P4 P7 P8 O1 O2 PO3 PO4 PO5 PO6 PO7 PO8 VV vs. VI <0.001 <0.001 <0.001 <0.001 0.045 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 IV vs. VI <0.001 <0.001 <0.001 0.006 <0.001 <0.001 VV vs. IV <0.001 0.004 0.002 In all cases Bonferroni correction was applied Table V Reaction times of the four experimental stimulus sequences analyzed in Experiment 2 Condition Reaction Times/ Mean SD Valid / Valid 346.37 86.38 Invalid / Valid 363.78 83.69 Invalid / Invalid 382.64 103.46 Valid / Invalid 401.99 87.81 Table VI Percentage of errors in the stimulus sequence of Experiment 2 Condition Percentage of no response SD Percentage/ S1 SD Percentage/ S2 SD Error Invalid 7.72 6.44 0.29 0.79 1.78 3.08 Valid 5.054 3.5 0.41 0.5 0.25 0.52 Omissions (percentage of no response), false alarms to the S1 of the second trial of the sequence (i.e. S1 before target, indicated as percentage/S1), and false alarms to the target stimuli (percentage/S2) are shown. 162 C.M. Gómez et al. (P<0.04). No other comparisons (responses to S1 and omissions to S2) were statistically significant. Comparison between Tables II and VI shows that the main difference between the two experiments was a greater number of omission errors in Experiment 2 than Experiment 1, probably because it was more difficult to discriminate the targets. In any case, any erroneous trial or sequence of trials was eliminated when the ERPs and RTs were computed. Event-Related Potentials P3a Figure 5 shows a greater positivity for the VI condition than for the VV and IV conditions. The waves are entirely similar to those depicted in Fig. 3 for Experiment 1. ANOVA showed an interaction between the type of trial sequence and the electrodes (F4,60=2.71, P=0.038). Posthoc analysis revealed effects at all the electrodes considered when the VI condition was compared to IV and VV (Table VII). There were no statistically significant differences between VV and IV at any of those electrodes. P3b ANOVA showed that the interaction between the type of trial sequence and electrodes was statistically significant (F4,60=7.51, P<0.001). Post-hoc comparisons revealed statistically significant differences between the IV and VV conditions at electrode P4 (Table VIII), the posterior positivity being greater in the IV than the VV condition. Figures 5 and 6 show a larger P3b in the IV than the VV condition at the posterior right side of the scalp. Post-hoc comparisons between VI and VV Table VII Post-hoc results for the component P3a in the Experiment 2. Fz Fcz Cz VV vs. VI <0.001 <0.001 <0.001 IV vs. VI <0.001 <0.001 <0.001 VV vs. IV In all cases Bonferroni correction was applied. The cells with no values were not statistically significant. Table VIII Post-hoc results for the component P3b in the Experiment 2 P3 P4 P7 P8 O1 O2 VV vs. VI <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 IV vs. VI <0.001 <0.001 <0.001 0.046 <0.001 <0.001 VV vs. IV 0.017 In all cases Bonferroni correction was applied Fig. 3. ERPs for the different sequences of trials in Experiment 1. Notice that the P3a component at the FCz electrode is greater for the valid–invalid condition than the valid–valid or invalid–valid conditions. The P3b is greater in the valid–invalid condition than in the valid–valid condition at the PO7 electrode, and greater in the invalid–valid than the valid–valid condition at the P8 electrode. The horizontal bar indicates the time window of the maps shown and the statistic referred to in the text. Trial-by-trial learning and P3 163 or IV were statistically significant at all electrodes considered. CNV during the S1–S2 period Further analysis was conducted to test the possibility that the S1–S2 preparatory period was modified, as indicated by the CNV, as a function of the outcome of the previous trial. Figure 7 shows the CNVs for trials preceded by an invalid or a valid trial in the two experiments. Although there were some differences in the CNV as a function of the outcome of the previous trial, those differences were not consistent across experiments and were not analyzed further. DISCUSSION Both experiments gave the same type of results: the RTs showed the trend VV < IV < II < VI. However, the VV condition was significantly faster than the IV condition only in the second experiment. In both experiments, P3a and P3b were larger in the VI condition than in VV or IV. However, the comparison most relevant to the sequential effects in these experiments was VV vs. IV. P3a showed no statistically significant differences between these two conditions. However, comparison of P3b in IV and VV showed statistically significant differences: it was larger to valid targets preceded by an invalid trial than to valid targets preceded by a valid trial. Therefore, the general trend for P3b amplitude was VV < IV < VI. These results suggest trial-by-trial learning of the predictive value of the cue. Behavioral results RTs were longer for invalidly cued targets (II and VI conditions) than for validly cued ones (VV and IV conditions). This is a general result for the central cue Posner’s paradigm (Posner 1980) and has been interpreted as a consequence of allocating spatial attentional resources in the pre-cued position, producing the typical cost-benefit pattern to invalid and valid targets. Both experiments showed more errors in the invalid conditions than valid ones. Although the two experiments showed statistically significant differences in only some RTs under the four different experimental conditions, the general pattern Fig. 4. ERPs (P8) and voltage maps of invalid–valid and valid–valid trials and the difference wave in Experiment 1. Notice the presence of a posteriorly-distributed P3b component under both IV and VV trial conditions. The difference wave shows that the amplitude of the P3b component is greater in the IV than the VV condition. The horizontal bar indicates the time window of the maps shown and the statistics referred to in the text. Fig. 5. ERPs for the different sequences of trials in Experiment 2. Notice that the P3a component at the FCz electrode is higher in the valid–invalid than the valid–valid and invalid–valid conditions. For the P3b at the P4 electrode, notice the gradual increase of amplitude from the valid–valid to the valid–invalid condition; the invalid–valid condition presents an intermediate value. The horizontal bar indicates the time window of the maps shown and the statistic referred to in the text.