The involvement of right dorsolateral prefrontal cortex (rDLPFC) in explicit temporal processing is well documented. Conversely, the role of this area in implicit temporal processing (e.g., foreperiod [FP] effect) is still poorly understood. The FP effect, usually observed when a range of variable FPs occur randomly and equiprobably, consists of reaction times (RTs) decreasing as the FP increases. Moreover, in such paradigms, RTs increase as a function of the preceding FP (i.e., sequential effects). Patients with lesions of the rDLPFC do not show the typical FP effect. The present study aimed to replicate these results in healthy adults using transcranial magnetic stimulation (TMS) and to further investigate whether any change of sequential effects follows a reduction of the FP effect. The results of 2 experiments (with simple and choice RT tasks, respectively) indicate that the FP effect was significantly reduced after TMS over the rDLPFC, whereas no effect was observed after stimulation of a left contralateral site and the right angular gyrus. Conversely, sequential effects were not influenced by TMS. A dual-process model of the FP phenomena is proposed to interpret the dissociation found between the 2 effects.
A number of models postulate the existence of an anterior attentional system with a set of supervisory functions (e.g., Norman and Shallice 1986; Posner and Petersen 1990). Recent evidence suggests a fractionation of these high-level functions within a network of different functional prefrontal areas (e.g., Shallice 2004; Stuss and others 2005). However, according to an opposite view, each prefrontal area has an undifferentiated and equipotential role as different cognitive demands usually involve common, or at least overlapping, prefrontal regions (e.g., Duncan and Owen 2000). The present study is intended to provide a specific contribution to this more general debate by investigating which area within the prefrontal cortex is involed in a basic cognitive capacity, namely preparation over time. Although preparation is a ubiquitous cognitive process, the way in which it evolves over time, how the cognitive system makes use of temporal information to optimize behavior, and which brain regions are engaged during preparation are all issues that still remain unresolved.
A typical paradigm employed to study the relationship between time and nonspecific preparation is the variable foreperiod (FP) paradigm. In a reaction time (RT) paradigm, the FP is the unfilled time interval between the warning stimulus (WS) and the stimulus requiring a response, the so-called imperative stimulus (IS). When FPs at a number of points within a fixed range occur randomly over trials but with the same a priori probability, RTs are longer for shorter FPs and shorter for middle and longer FPs. (Strictly speaking, the FP is a continuous variable. In the current study, we decided to treat the FP as a categorical variable, using only 3 values of the FP [i.e., 0.5, 1, and 1.5 s]. This choice was aimed to allow comparison with the existing literature on the FP effect, where this variable has been almost always treated as categorical [for a review, see Niemi and Näätänen 1981], except for a few exceptions [e.g., Drazin 1961]). This is the variable FP effect (Woodrow 1914; for a review, see Niemi and Näätänen 1981). The traditional explanations of this phenomenon have attributed it to strategic processes. On these approaches, the FP effect is explained by assuming the existence of a process which continually checks the conditional probability of IS occurrence. This process leads to an increase over time in the strategically controlled state of preparation, if the IS has failed to occur so far during the FP (e.g., Drazin 1961; Näätänen 1970).
Moreover, sequential effects are also usually observed with this paradigm: longer FPs on the preceding trial produce slower RTs on the current one (Karlin 1959; Niemi and Näätänen 1981). This is particularly the case for the shortest current FPs. Hence, the sequential effects usually found are asymmetrically biased toward the shortest FPs in the range used. Traditionally, the sequential effects are explained by assuming that the cognitive system reaches peak preparation strategically at the same FP as the previous trial (Alegria 1975). Additionally, when a FP on the current trial is longer than on the previous one, it is assumed that participants can voluntarily extend the period of optimal preparation (Karlin 1959; Thomas 1967) or cyclically reprepare (Baumeister and Joubert 1969; Alegria 1975). This repreparation/maintenance hypothesis accounts for the asymmetry, in which sequential effects occur on the “long–short” FP sequences but not on the “short–long” ones.
However, recent empirical evidence presents a challenge to the strategic accounts. Los and van den Heuvel (2001) designed a temporal cueing paradigm with a variable FP design. In their Experiment 2, they used valid (80%) and invalid (20%) signals cueing the FP interval. With a valid cue, participants could intentionally prepare for the cued FP and, consequently, the sequential effects were drastically attenuated. The most impressive results, however, were obtained with an invalid cue for the FP duration. In particular, when an invalid cue for a “long” FP preceded an actual current “short” FP, the costs of not having prepared in advance for that short FP are not constant. They vary as a function of the FP that occurred on the preceding trial. Sequential effects are stronger as compared with those occurring after a validly cued FP. If sequential effects were fully strategic in nature, as traditional accounts claim, there would be no reason for them to influence costs after invalid cues. For these reasons, Los and van den Heuvel (2001; see also Los and others 2001) proposed an alternative account, in which both the FP and the sequential effects are integrated in a common theoretical framework based on conditioning mechanisms. For each FP, a level of activation is adjusted in a way that obeys the learning rules of trace conditioning (Machado 1997). It is assumed that, if any FP occurs on trial n − 1, then the conditioned strength of the representation of each FP in the range will be modified so that it will be 1) increased for that FP, 2) not influenced for longer FPs, and 3) decreased for shorter FPs because of a need to avoid anticipation responses. Thus, if an equally long FP occurs on trial n, the corresponding RT will tend to be faster than if a shorter FP had occurred. This conditioning view has the advantage of linking the FP effect closely to the sequential effects in a single-process account because RT on trial n is a function of the conditioning influences produced on trial n − 1.
On the other hand, the involvement of strategic control in a variable FP paradigm was indirectly supported by neuropsychological findings. As shown in a recent neuropsychological study (Stuss and others 2005), patients with lesions in the right dorsolateral prefrontal cortex (rDLPFC) fail to show the typical FP effect, unlike both control participants and other subgroups of prefrontal patients. Stuss and others (2005) suggest that the rDLPFC is likely to be the region responsible for the strategic process producing the FP effect, which controls the state of preparedness by checking the conditional probability of IS occurrence. This account fits a range of studies that attribute a monitoring role to the rDLPFC (Coull and others 1998; for reviews, see Fletcher and Henson 2001; Shallice 2004). Unfortunately, the sequential effects were not considered in the study of Stuss and others (2005). Hence, without an additional investigation of the sequential effects produced by the patients, the lack of FP effect shown can alternatively be interpreted as a failure in the conditioning processes, according to the conditioning view (Los and van den Heuvel 2001).
The aim of the present study is 2-fold: concerning both the anatomical and the functional bases of the FP phenomena. From an anatomical point of view, the aim was to investigate whether the neuropsychological results (Stuss and others 2005) could be reproduced in healthy adults by means of transcranial magnetic stimulation (TMS). In addition, however, the study could allow one to establish the role of the rDLPFC not only in the FP effect but also with respect to the sequential effects, which have not been investigated neuropsychologically. Moreover, the role of the right angular gyrus (AG) in the FP phenomena was investigated here because this area is often associated with temporal processing found in functional magnetic resonance imaging (fMRI) and TMS studies (Coull and others 1998; Rao and others 2001; Lewis and Miall 2003; Alexander and others 2005). It can be argued that these studies utilize tasks that require explicit time processing, whereas the time processing required in an FP paradigm is implicit. However, a recent study has also demonstrated anticipatory activity revealing an internal representation of both elapsed time and the probability of the stimulus occurrence in the macaque parietal lobe during an implicit time-processing task similar to the variable FP paradigm (Janssen and Shadlen 2005; see also Onoe and others 2001).
From the functional point of view, TMS can provide a strong test of the various cognitive models of the FP phenomena. If both the FP and the sequential effects are influenced by the TMS on rDLPFC, a common mechanism is likely to underlie them, as predicted by the conditioning view (Los and van den Heuvel 2001). If instead the FP effect only is modified, this would be evidence for a dual-process theory of the FP phenomena. In order to test these possibilities, 2 experiments were designed with a simple and a choice RT task (Experiments 1 and 2, respectively) and a theta burst stimulation (TBS) off-line paradigm of repetitive TMS, which has been demonstrated to transiently decrease cortical activity in the stimulated area (Huang and others 2005). Moreover, the range of FPs used in the study of Stuss and others (2005) (i.e., 3–7 s) might suggest the alternative explanation of a deficit of the vigilance system in maintaining attention over a long period of time, instead of a monitoring process, in the case of right prefrontal lesions. For this reason, a much shorter FP range was used (i.e., 0.5, 1, and 1.5 s) to discard vigilance-related accounts of a possible modulation of the FP effect.
In Experiment 1, a simple RT task was used in order to study the FP phenomena in their simplest form. For this purpose, a range of FPs (i.e., 0.5, 1.0, and 1.5 s) were employed in a standard variable FP paradigm. A repetitive TMS paradigm was used to investigate the role of the rDLPFC in the FP phenomena. In order to control the specificity of any effect obtained to the rDLPFC, 2 additional areas were stimulated during different sessions, namely, a site in the left DLPFC and another in the right AG.
Nine volunteer participants, 3 females and 6 males, took part in the experiment. They were 31 years old on average (range = 24–43 years). All of them were right handed with an average score of 92 on the Edinburgh Handedness Inventory (Oldfield 1971). All had normal or corrected-to-normal vision and no auditory or neurological impairment. This study was approved by the UCL Committee on the Ethics of Human Research.
Apparatus and Materials
The experiment was conducted on a PC. Stimuli were presented on a 19-inch monitor with a 100-Hz refresh rate. Participants viewed the display at a distance of about 60 cm from the center of the monitor, with the index finger of their dominant hand resting on the keyboard spacebar. The auditory WS was a 1500-Hz pure tone, presented for 50 ms through bilaterally located speakers. All visual stimuli were presented on a black background. A centrally presented cross, consisting of 2 yellow crossed bars 1.0 × 0.5 cm in size, served as a fixation stimulus. The IS was a downward pointing white arrow 2 cm long, which consisted of a 1.5 × 1–cm bar attached to a 0.5-cm arrowhead with a maximum width of 2 cm. Two blocks were presented during each session to each participant, one before and the other after TMS. In each block, 3 FPs of 0.5, 1.0, and 1.5 s, respectively, were presented for an equal number of trials (i.e., 48 trials each), randomly drawn from a rectangular probability distribution. Each FP was preceded with the same probability by a FP of 0.5, 1.0, and 1.5 s.
Procedure and Task
Figure 1 displays the hierarchical organization of Experiment 1. The experiment consisted of 3 sessions performed in 3 different days within 1 week, one for each area stimulated (see below). Each session consisted of 2 blocks of 144 trials each. The first block was run before the TMS to measure the baseline performance. The second block was run after the TMS to measure its effects on the behavioral performance. Unpublished data from our laboratory show that the learning period for the FP phenomena, if any, is very brief as after an initial block of 60 trials, the FP and sequential effects are already present with the same magnitude as in subsequent blocks. For this reason, we could be confident that the performance of the baseline block would not influence the performance on the post-TMS block. Participants were tested individually in a silent and dimly lit room. They received written instructions explaining the simple RT task. A trial started with the presentation of the fixation cross together with the WS. The onset of WS marked the beginning of the FP. When the FP ended, the fixation cross disappeared, and the IS was presented at the center of the screen. Participants were instructed to respond as fast as possible to the IS by pressing the spacebar with the index finger of their right hand. The IS was removed by the response key press or after a deadline of 1.5 s. After a blank interval of 1 s, a new trial started. Nine practice trials were given at the beginning of the test. An experimental session lasted about 25 min each day (i.e., 6–7 min per block plus some minutes after the baseline block needed to find the site coordinates in the brain).
Locations for TMS were determined using the Brainsight TMS–magnetic resonance imaging coregistration system (Rogue Research, Montreal, Canada), through conversion of the Montreal Neurological Institute (MNI) stereotaxic coordinates to participant's normalized brain using the software SPM2. For each session, participants underwent the coregistration in the interval between the two pre- and post-TMS blocks, in order to find the coordinates of the area of interest in their real brain. Those coordinates were taken from an fMRI study by Lewis and Miall (2003), in which activation of the right and left DLPFC and of the right AG was found during a temporal discrimination task for both sub- and suprasecond intervals. For each participant and session, one area was stimulated after the performance of the pre-TMS baseline block. The areas to be stimulated were located in the rDLPFC (MNI stereotaxic coordinates: 48, 42, 24), in a control area on the left DLPFC (the mirrored contralateral site, i.e., −48, 42, 24), and in the right AG (48, −45, 48), respectively (see Fig. 2). The order in which the 3 areas were stimulated was counterbalanced across participants, such that each area was stimulated in the first day session on an equal number of participants.
Participants wore a latex swimming cap on which the location found in the coregistration procedure was marked with a phosphorescent spot. Stimulation was produced through a MagStim Super Rapid stimulator with 4 external boosters with a maximum output of approximately 2 T (MagStim, Whitland, UK). A figure-of-eight 50-mm coil was used for the stimulation with the center of the coil positioned over the marked spot such that the windings were at 90° to the scalp and the handle pointed vertically.
An off-line TMS paradigm was chosen rather than an online one to prevent any exogenous influence of the sound and the proprioceptive sensation given by the TMS on the RTs (e.g., Terao and others 1997) and hence on the FP phenomena. In each session, the TMS parameters were those of the continuous TBS, consisting of a burst of 3 pulses at 50 Hz (i.e., 20 ms among each stimulus), which was repeated at intervals of 200 ms for 20 s (giving a total of 300 pulses). The output strength of the TMS was set to 80% of the participant's active motor threshold, defined as the minimal intensity of stimulation capable of inducing a visible twitch of the contralateral first dorsal interosseus in at least 6 trials out of 10, by means of a single pulse delivered at the best scalp position over motor cortex. During the calculation of the active motor threshold, the participant's dominant hand was in a moderately contracted position and the thumb and index fingers were in opposition. Previous studies (Di Lazzaro and others 2005; Huang and others 2005) have demonstrated that this TBS protocol temporarily produces reduced excitability of motor cortex outlasting the period of actual TMS. With 20 s stimulation, the time window of reduced excitability was estimated to last up to 20 min.
Trials on which the RT was outside the 100–1000 ms range and trials where anticipated responses were made (viz. ones occurring before the IS) were discarded from further analyses. In addition, the first trial of each block was eliminated. Mean RTs for each participant and condition were analyzed by repeated-measures analyses of variance (ANOVAs).
As a first step, 3 separate repeated-measures ANOVAs have been conducted contrasting baseline against post-TMS blocks separately for each site stimulated during each session (rDLPFC, left DLPFC, right AG). The within-subject independent variables considered in these preliminary analyses included FP on the current trial (FPn = 0.5, 1.0, and 1.5 s), FP on the preceding trial n − 1 (FPn − 1), and the TMS block (Baseline vs. post-TMS). The dependent measure was the mean RT.
In order to investigate differences across TMS sessions, a subsequent overall repeated-measures ANOVA was then performed with FPn, FPn − 1, and TMS site (rDLPFC, left DLPFC, and right AG) as the within-subjects variables. The dependent variable chosen in this overall ANOVA was the degree of change in mean RTs as a result of stimulation at each site compared with the mean RTs of the pre-TMS baseline block of the same session (with 100% representing no RT change, a value >100% representing a slowing down, and one <100% indicating a speeding up) (We chose this dependent variable as it directly estimates the direction of any change of the post-TMS RTs with respect to the baseline RTs [i.e., slowing or speeding]. However, alternative ANOVAs were also conducted on the raw RTs, with FPn, FPn − 1, TMS site [right and left DLPFC and right AG], and block [baseline and post-TMS block in Experiment 1; baseline, first and second post-TMS blocks in Experiment 2]. These ANOVAs gave exactly the same results as those reported here, in both experiments. Specifically, the FPn × TMS site × block interaction was significant [F4,32 = 3.4, P < 0.05; F8,64 = 2.3, P < 0.05 for the Experiments 1 and 2, respectively], confirming the selective reduction of the FP effect after stimulation of the rDLPFC).
For the significant effects, post hoc Tukey's honestly significant difference (HSD) comparisons were performed, in order to see which comparisons accounted for the effects. An effect was defined as significant if its corresponding α-level was below 0.05. The F-test was adjusted by the Greenhouse–Geisser ϵ correction when the Mauchley sphericity test was significant.
The first trial of each block (0.7%), trials involving anticipated responses (RT before IS onset: 1.9%), trials involving premature responses (RT < 100 ms: 0.5%), and trials with delayed responses (RT > 1000 ms) or without responses (0.03%) were discarded from further analyses. Fewer than 3.2% of trials were excluded with no difference across conditions.
The FP and sequential results were produced as expected. The following effects were significant in ANOVAs conducted on the 3 sessions separately (see Fig. 3). First, the main effect of FPn was obtained (F2,16 = 28.7, P < 0.001; F1.1,9.1 = 40.1, adjusted P < 0.001; F2,16 = 34.1, P < 0.001 for rDLPFC, left DLPFC, and right AG, respectively). Post hoc Tukey's HSD comparisons showed that RTs were slowest for the shortest FP of 0.5 s as compared with the middle and longest FPs (for all, P < 0.001), but there was no difference among the RTs for FPs of 1.0 and 1.5 s in any session. The main effect of FPn − 1 also reached significance (F1.1,8.8 = 5.5, Adj. P < 0.05; F1.2,9.9 = 15.3, Adj. P < 0.01; F1.3,10.1 = 17.2, Adj. P = 0.001 for rDLPFC, left DLPFC, and right AG, respectively). The post hoc comparisons showed that current RTs were slower following a longest FPn − 1 trial than a shortest FPn − 1 trial (for all, P ≤ 0.01). The FPn × FPn − 1 interaction, concerning asymmetry of sequential effects, was also significant (F4,32 = 23.6, P < 0.001; F4,32 = 11.4, P < 0.001; F2.4,19.6 = 6.5, Adj. P < 0.01 for rDLPFC, left DLPFC, and right AG, respectively). This effect was primarily due to the differential contribution of the FPn − 1 to the RT at each of the 3 current FPs, being greatest for the shortest FPn and smallest, virtually absent, for the longest FPn, as confirmed by post hoc comparisons.
TMS Effects: Separated ANOVAs for Each TMS Site
No TMS block main effect was observed for any site of stimulation, indicating that RTs were not nonspecifically modified by the TBS. More critically, the TMS block × FPn interaction was significant only for the rDLPFC session (F2,16 = 7.2, P < 0.01), indicating that the FP effect was reduced in the post-TMS block as compared with the baseline block during the rDLPFC session. This conclusion was corroborated contrasting the RT difference between the shortest and the longest FPn in the baseline block (43 ms) and the same RT difference in the post-TMS (26 ms) by means of planned comparisons (P < 0.001). In addition, there was a tendency for a TMS × FPn − 1 interaction in the rDLPFC session (F2,16 = 3.5, P = 0.054), suggesting that the effect of the FPn − 1 was slightly reduced after TMS with respect to the baseline block. However, the TMS block × FPn × FPn − 1 3-way interaction was not significant in any session (for all, P > 0.3), indicating that the sequential effects were not significantly modulated by the TMS.
TMS Effects: Overall ANOVA
The only significant effect obtained in the overall ANOVA was the interaction between TMS site and FPn (F4,32 = 2.8, P < 0.05), indicating that the FP effect was selectively modulated after TMS over the rDLPFC (see Fig. 4). In this case only, indeed, the RTs on the longest FPn were slower with respect to the baseline (104%), whereas the RTs on the shortest FPn were faster (97%), the difference between these 2 values being significant following post hoc Tukey's comparisons (P < 0.01). It is noteworthy that this modulation goes in the opposite direction with respect to the FP effect itself, slowing RTs down on the longest FPn and speeding them up on the shortest FPn. Notably, no TMS modulation was found for the sequential effects, as indicated by the lack of significance for the TMS site × FPn × FPn − 1 3-way interaction (P = 0.45).
The FP effect and the asymmetric sequential effects, usually found in a variable FP paradigm, were replicated in Experiment 1. Moreover, a role of the rDLPFC was demonstrated for the FP effect. Following TMS over the left DLPFC and the right AG, no significant changes were observed with respect to the baseline. On the other hand, after TMS over the rDLPFC, a reduction in the FP effect was observed, confirming the role of this area in the occurrence of this effect, as already shown by a recent neuropsychological study (Stuss and others 2005). This result is interesting per se but cannot be used to corroborate any dual-process model of the FP phenomena. A dissociation between the FP and the sequential effects was not clearly obtained in this experiment because the effect of the FPn − 1 also showed a tendency toward being reduced after the TMS over the rDLPFC.
As the reduction of the cortical excitability of the motor cortex reaches a maximum from 7 to 14 min after TBS (Di Lazzaro and others 2005; Huang and others 2005), a similar time course can be expected in the case of the rDLPFC. For this reason, in the second experiment, 2 blocks of trials were performed by participants in order to cover, apart from an initial 6-min period similar to Experiment 1, an additional subsequent period when the stimulation effects are supposed to be at their greatest.
Experiment 2 was designed to be similar to Experiment 1, with 2 exceptions. First, there were 2 blocks of trials instead of one after the TMS. This change was made in order to cover the whole period in which the long-lasting effects of TBS operate, including the most critical interval of 7–14 min post-TMS, possibly increasing the sensitivity of experiment. Second, a 2-choice RT paradigm was chosen in order to avoid anticipation and ceiling effects that might occur during a simple RT task, such as that used in Experiment 1; such effects could have partially obscured any TMS effects. It should be noted that, although the additional stage of response selection is required by a choice RT task, the FP effect is held to involve “nonspecific” preparation processes only. Indeed, the typical FP phenomena are basically similar for both simple and choice RT tasks (Los and van den Heuvel 2001), and the right prefrontal lesion effect is found for both paradigms (Stuss and others 2005).
Nine volunteer participants, 4 females and 5 males, took part in Experiment 2. They were 30 years old on average (range = 22–43 years). All of them were right handed (with an average score of 83 at the Edinburgh Handedness Inventory). All had normal or corrected-to-normal vision and no auditory or neurological impairment. Three of them had participated in Experiment 1 (at least 1 month before).
Apparatus and Materials
The apparatus and materials were the same as in Experiment 1 except that, on each trial, the IS appeared with the same probability to the right or left of the fixation cross at a distance of 7.5 cm (center to center).
Procedure and Task
The task was to perform a spatially compatible response to the IS by pressing a right or left key on the computer keyboard (“F” or “J”, respectively) as fast as possible, with the left and right index fingers, respectively. The experiment consisted of 3 sessions performed on 3 different days within a maximum period of a week. Each session consisted of 3 blocks of 144 trials each. During each session, the first block was run before the TMS to measure the baseline performance and the second and third blocks were run consecutively after the TMS. The whole experimental session lasted about 35–40 min.
The same criteria as in Experiment 1 were used for the analysis of the data obtained in Experiment 2. In addition, only trials with correct (spatially compatible) responses to the IS were considered for the RT analyses. Moreover, preliminary analyses including the spatial position of the IS (and of the response) did not give rise to any interaction between this factor and that of the TMS site factor. For this reason, the spatial position factor was collapsed in the following analyses.
As a first step, 3 separate repeated-measures ANOVAs were conducted, one for each stimulated site (rDLPFC, left DLPFC, and right AG). The within-subject independent variables considered in these preliminary analyses included FP on the current trial (FPn = 0.5, 1.0, and 1.5 s), FP on the preceding trial n − 1 (FPn − 1), and the TMS block (Baseline, first and second post-TMS blocks). The dependent measure was the mean RT.
A subsequent overall repeated-measures 3 × 3 × 2 × 3 ANOVA was also performed with FPn, FPn − 1, TMS block (first and second post-TMS blocks), and TMS site (rDLPFC, left DLPFC, right AG) as the within-subjects variables. The dependent variable chosen in this overall ANOVA was the percentage of change of mean RTs collected after the first and second post-TMS blocks in each session compared with the mean RTs of the pre-TMS baseline of the same session. As in Experiment 1, the F-test was adjusted by the Greenhouse–Geisser procedure when appropriate.
The first trial of each block (0.7%), trials involving anticipated responses (RT before IS onset: 0.08%), trials involving premature responses (RT < 100 ms: 0.02%), trials with delayed responses (RT > 1000 ms) or without responses (0.02%), and trials with an incorrect response on the spatial compatibility task (0.8%) were discarded from further analyses. Fewer than 1.8% of trials were excluded.
The following effects were significant in ANOVAs conducted on the 3 sessions separately (see Fig. 5). A main effect of FPn was obtained (F2,16 = 10.1, P = 0.001; F1.3,10.1 = 15.8, Adj. P < 0.01; F2,16 = 17.3, P < 0.001 for the rDLPFC, left DLPFC, and right AG, respectively). Post hoc comparisons showed that RTs were slowest for the shortest FP of 0.5 s as compared with the middle and longest FPs (for all, P < 0.01). Moreover, the main effect of FPn − 1 was also significant (F2,16 = 16, P < 0.001). The post hoc comparisons showed that current RTs were slower following a middle FPn − 1 trial than a shortest FPn − 1 trial and, in turn, following a longest FPn − 1 trial than a middle one (for the right AG, the difference between the middle FPn − 1 and the longest one was not significant, for all the other comparisons, P < 0.05). The FPn × FPn − 1 interaction, concerning the asymmetric sequential effects, was also significant (F1.8,14.8 = 5.1, Adj. P = 0.02; F4,32 = 12.7, P < 0.001 for the rDLPFC and left DLPFC, respectively; and a tendency for the right AG, F1.5, 12 = 3.9, Adj. P = 0.059). This effect was principally due to the differential contribution of the FPn − 1 to the RT on each of the 3 current FPs, being greatest for the shortest FPn and smallest for the longest FPn, as confirmed by post hoc comparisons.
TMS Effects: Separated ANOVAs for Each TMS Site
No TMS block main effect was observed for any site of stimulation, indicating that RTs were not nonspecifically modified by the TBS. As for Experiment 1, the TMS block × FPn interaction was significant for the rDLPFC session only (F4,32 = 4.5, P < 0.01). The FP effect was reduced in the first and second post-TMS block with respect to the baseline, as demonstrated by subsequent planned comparisons. Significant planned comparisons were obtained contrasting the RT difference between the shortest and the longest FPs in the baseline condition and the same RT difference in the first and in the second blocks after TMS of the rDLPFC (for all, P < 0.001). However, post hoc comparisons indicated that the differences among RTs in any of the FPs in the baseline contrasted with the same FPs in the post-TMS blocks were not significant (for all, P > 0.1). As in Experiment 1, this pattern indicates that the reduction of the FP effect was not due to a specific effect on RTs for the shortest, middle, or longest FP but instead to an overall effect that acts by attenuating the FP effect as a whole. Thus, the difference between the RT on the shortest FPn and on the longest FPn decreased from 33 ms in the baseline block to 22 ms in the first post-TMS block, to reach 15 ms in the last post-TMS block. No other effect was significant. In particular, a lack for a 3-way TMS block × FPn × FPn − 1 interaction in all the 3 sessions (for the left DLPFC, P > 0.8; for the right AG, P > 0.6; for the rDLPFC, P > 0.2) suggested that TMS did not modulate sequential effects.
Although no 3-way interaction was observed in any analysis, visual inspection of Figure 5 suggests, for the rDLPFC session, the presence of more symmetric sequential effects in the last TMS block with respect to the baseline. The 3-way interaction may be hidden by the number of conditions introduced in the ANOVA. To statistically assess this possibility, a subsequent 3 × 3 × 2 ANOVA was performed for the rDLPFC session, with FPn, FPn − 1, and TMS block as within-subjects variables, contrasting the baseline block with the second TMS block only. This analysis was justified by the fact that the TMS effect was expected to reach a maximum during this second block (i.e., from 7 to 14 min post-TMS; Huang and others 2005). Apart from the main effects of FPn and FPn − 1 and the interaction between these 2 factors, which confirm the previous analysis, this analysis produced a significant 3-way interaction (F4,32 = 2.7, P < 0.05). To confirm that the difference was due to the presence of symmetric sequential effects in the second post-TMS block, further planned comparisons have been carried out. Specifically, these comparisons contrasted the difference between the RT on the longest FPn preceded by the longest FPn − 1 and the RT on the same longest FPn preceded by the shortest FPn − 1 in the baseline with the same difference in the second post-TMS block. This analysis was significant (F1,8 = 13.9, P < 0.01), confirming that, in the second post-TMS block, sequential effects were basically symmetric, as they were also present on the longest FPn.
TMS Effects: Overall ANOVA
The main effect of the FPn was significant (F2,16 = 5.1, P < 0.05). Post hoc comparisons indicated that post-TMS RTs were slower after a long FPn and faster after a short FPn with respect to the baseline values (101% vs. 99%; P < 0.05). More importantly, an interaction between TMS site and FPn was obtained (F4,32 = 3.5, P < 0.05), indicating that the FP effect was selectively attenuated after TMS on the rDLPFC (see Fig. 6). In this case only, the RTs on the longest FPn were slower with respect to the baseline block (103%), whereas the RTs on the shortest FPn were slightly faster (99%), the difference between these 2 values being significant on post hoc Tukey's comparisons (P < 0.05). This effect demonstrates a reduction of the FP effect selectively after the TMS of the rDLPFC, replicating the results of Experiment 1. No other effect was significant. As in the first experiment, the left DLPFC and the right AG turned out to act as TMS control sites because no effect involving TMS was observed in these analyses.
The results of Experiment 2 confirm those of Experiment 1. In particular, the reduction of the FP effect as a consequence of TMS acting specifically on the rDLPFC was replicated, whereas no TMS effect was obtained on the left DLPFC and the right AG. This result confirms the role of the rDLPFC for the occurrence of the FP effect (see Stuss and others 2005).
On the other hand, the sequential effects were not reduced by the TMS in the rDLPFC, clearly suggesting that the nonsignificant tendency found in Experiment 1 was likely to be due to noise. As dissociation between the 2 effects has been obtained, the hypothesis of a common mechanism underlying both effects is not supported. Interestingly, the sequential effects become more symmetric during the second post-TMS block of trials on the same area. This was confirmed by the presence of a significant FPn × FPn − 1 × TMS block interaction, although this was obtained only for a direct comparison between the baseline block and the second post-TMS one, namely, the block in which the effects of the TBS would be expected to be stronger (see Huang and others 2005). Such evidence, although not confirmed by the analysis across TMS sessions, corroborates the pattern found in a developmental study of the FP phenomena by A. Vallesi and T. Shallice (in preparation, see General Discussion). As in the youngest children of that study (i.e., 4–5 years old), whose DLPFC is presumably not yet mature, after 6–7 min from TMS of the rDLPFC, the sequential effects become more symmetric, whereas the FP effect decreases.
The FP effect can be briefly described as a negatively accelerating RT–FP function, obtained when a range of FPs is randomly administered with the same a priori probability in simple or choice RT tasks. The present study was primarily designed to investigate the anatomical basis of the FP effect by means of TMS. To this purpose, in Experiment 1 a variable FP paradigm with a simple RT task was performed by 9 volunteers in 2 blocks, before and after TMS of 3 sites, rDLPFC, left DLPFC, and right AG (one per each session day). A possible limitation of Experiment 1 was that testing did not take place in the most critical period of the TMS effect when the TBS is employed (i.e., 7–14 min, see Huang and others 2005). This limitation was overcome in the second experiment, where 9 participants carried out 2 experimental blocks following TMS. Stuss and others (2005) found a right frontal lesion effect on variable FP paradigms using choice RTs as well as simple RTs. So, another change introduced in the second experiment was the use of a choice RT task instead of the simple RT task used in Experiment 1.
Both experiments of the present study provide clear evidence for a role of the rDLPFC in the production of the FP effect. Specifically, a 20 s off-line repetitive TMS (i.e., TBS) over this area, compared with the pre-TMS baseline level and to an analogous stimulation on 2 other sites (i.e., left DLPFC and right AG), is enough to significantly reduce the size of the FP effect for the post-TMS period investigated. This period was about 6 min long for the simple RT task in Experiment 1 and 15 min long for the choice RT task in Experiment 2. These results fit well with existing neuropsychological literature. As suggested by Stuss and others (2005), the rDLPFC seems to be the location of a process critical for the FP effect because patients with lesions of this region do not show the typical FP effect. Moreover, a link between the FP effect and dopaminergic activity has been demonstrated in neuropsychological studies. The dopaminergic system is a neurotransmitter system massively present within the DLPFC. Drug-free schizophrenic patients, for example, who have increased levels of dopamine (see Kapur 2003, for a review), show an exaggerated variable FP effect (Zahn and others 1963), whereas Parkinsonian patients, who suffer from deficiencies in dopamine levels (e.g., Rakshi and others 1999), have a reduction of this effect (Jurkowski and others 2005).
The short FP range used in the present study (i.e., 0.5, 1, and 1.5 s) allows us to discard accounts relating the reduction of FP effect after TMS of the rDLPFC to vigilance or alertness (e.g., see Posner and Petersen 1990), such as a deficit in maintaining a high-level of preparation for a long time interval. This account would be also in contrast to the results obtained by Stuss and others (2005) on prefrontal patients. In that study, indeed, right frontal patients showed no FP effect with a variable FP paradigm. Noteworthy, with a fixed FP paradigm, their performance was comparable, even for the longest FPs in the range, with that of other frontal patients, apart from medial frontal ones.
The FP effect is probably due to a process of endogenous preparation, analogous to that already described by some traditional accounts (e.g., Näätänen 1970). This process checks the information for the nonoccurrence of the IS, using the information derived from the increasing conditional probability of the IS occurrence as time elapses in order to enhance preparation. As in rDLPFC patients (Stuss and others 2005), it is likely that this checking process does not operate efficiently after TMS over this area. The rDLPFC region controlling such process would be permanently damaged in the neuropsychological patients studied by Stuss and others (2005) and compromised transiently in the present study using TMS.
In partial support of this view, activation of the rDLPFC has already been demonstrated in a number of brain imaging and TMS studies during tasks dealing with temporal processing, such as time discrimination tasks (e.g., Rao and others 2001; Lewis and Miall 2003) and time reproduction tasks (e.g., Basso and others 2003; Koch and others 2003; Jones and others 2004). It can be argued that these tasks are different in nature from the variable FP task, in that the former overtly require temporal processing, whereas the latter does not. However, what the 2 kinds of tasks may have in common is the monitoring of temporal information, which is explicitly required by time reproduction and discrimination tasks and implicitly exploited during the variable FP paradigm in order to reach an optimal level of preparation (Näätänen 1970). As the results of the present study and of the literature empirically suggest, this monitoring process is conceivably subserved by the rDLPFC.
The present study also investigated the role of the rDLPFC in the sequential FP effects. These well-known effects consist of an increase in RT as the preceding FP becomes longer. The sequential effects are usually asymmetrically more pronounced for the shortest FPs in the range employed, whereas they are virtually absent for the longest FPs. A tendency (P = 0.054) for a reduction of the effect of FPn − 1 was found after stimulation of the rDLPFC in the first experiment that used a simple RT task. This tendency was, however, no different following TMS of rDLPFC from that of the other 2 sites employed (see results of the overall ANOVA). Moreover, in Experiment 2, a dissociation between the FP effect and the sequential effects was obtained after stimulating the rDLPFC: the FP effect was significantly reduced, whereas the sequential effects were not influenced in magnitude. Thus, we obtained site controls (i.e., rDLPFC vs. all other sites and no TMS) and a task control for our effects (i.e., FP vs. sequential effects dissociation).
Moreover, in the second block after TMS of the rDLPFC, when the TMS effects are supposed to be at their strongest (Huang and others 2005), the sequential effects were more symmetric with respect to the baseline block, as confirmed by a subsequent ANOVA. Symmetric sequential effects would not be expected according to the conditioning view (Los and van den Heuvel 2001). This account predicts that the sequential effects, if present, must inevitably be asymmetric, due to a main role of extinction on the activation strength of the shortest FPs and of reinforcement on the activation strength of longest FPs. The observation of symmetric sequential effects when the FP is attenuated may be tentatively accounted for by a recently proposed dual-process model (A. Vallesi and T. Shallice, in preparation). According to this model, the sequential effects per se can be explained by assuming an enhancement in arousal following a short FPn − 1 and a decrease following a long FPn − 1, whatever the current FPn. In support of this view, tonic influence of the FPn − 1 on preparation was electrophysiologically tracked in a recent study by Los and Heslenfeld (2005). In that study, the contingent negative variation (CNV) was used as a covert index of nonspecific preparation, during a variable FP paradigm employing temporal cuing. The results show phasic effects of cuing on the CNV amplitude at the end of the short FPn. In that case, the CNV was more negative following signals cueing a short FPn (valid cue) than a long FPn (invalid cue). Conversely and more relevant for our purposes, the FPn − 1 influenced the CNV in a tonic fashion, independently of the cue. In particular, the amplitude of the ongoing CNV was lower throughout the FPn with a long FPn − 1 than with a short one. Notably, this pattern was observed “symmetrically” for both short current FPn and long current FPn. It is likely that sequential effects are usually compensated for during long FPs by the processes underlying the FP effect, thus accounting for their asymmetry. However, given the lack of a significant interaction between the sequential effects and the TMS site, further investigation is required to test the relationship between FP effect size and degree of symmetry of the sequential effects suggested by the pattern of the present results.
To the best of our knowledge, this is the first study in which an effect of the TBS parameters was demonstrated on behavioral variables (i.e., RTs) not directly linked to the measure of the motor cortex excitability, indicating its usefulness in cognitive neuroscience, as a feasible alternative to other kinds of off-line stimulation.
In conclusion, the present study demonstrated that the rDLPFC is a critical locus of the FP effect, as a reduction in this effect is obtained after TMS over that area. Moreover, the dissociation between the FP effect and the sequential effects produced by TMS of the rDLPFC, especially in Experiment 2, suggests a dual-process account, according to which the 2 (usually interacting) effects are likely to be dissociable both functionally and anatomically.
This research was supported in part by a grant from Programmi di Ricerca di Rilevante Interesse Nazionale to TS and Raffaella Rumiati and by grants from the Royal Society, the Wellcome Trust, and the Leverhulme Trust to VW. Conflict of Interest: None declared.