Serial Organization of Human Behavior in the Inferior Parietal Cortex

Discussion

The present results first reveal that all posterior cortical activations related to the execution of prelearned behavioral sequences were confined to the inferior parietal cortex, confirming the prominent role of this region in processing behavioral sequences. Additionally, all these activations appear independent of the nature of processed stimuli (squares vs letters in the motor and task condition, respectively), because all reported regions exhibited phasic activations in both the motor and task conditions. Similarly, increased phasic activations at sequence boundaries are unlikely to reflect selective processing of the perceptual content of start and stop cues (colors). Indeed, regions showing such activations also exhibited increased activations in sequential increment compared with baseline trials or did not respond to specific sequence boundary trials (task sequence boundaries in the right IPL and exogenous termination in the SMG/TPJ region). Thus, the data provide evidence that the reported inferior parietal activations are unrelated to the processing of the perceptual content of behavioral sequences.

The results especially reveal a single bilateral region located in the posterior IPS exhibiting sustained activations during the execution of both motor and task sequences. We did not find any other regions in the posterior cerebral cortex, and we did not previously identify any frontal regions showing the same activation pattern (Koechlin and Jubault, 2006). In the left IPS, the region also exhibited the same pattern of phasic responses during the execution of both motor and task sequences with especially increased activations at sequence boundaries. Such a joint pattern of sustained and phasic activations is expected in a region processing the abstract serial structure of action sequences (see Materials and Methods, Predictions), regardless of their hierarchical organization and sensorimotor content. According to this interpretation, sustained activations correspond to the maintenance of the serial structure and possibly the current position within this serial structure that guide sequential progression along behavioral sequences. Consistently, increased phasic activations at sequence boundaries simply reflect activation and inhibition of the entire sequence representation, whereas phasic activations associated with sequential increments correspond to updating the current position within such a representation.

Critically, the combination of sustained and phasic activations in this region rules out several alternative interpretations. Sustained activations cannot be simply explained by the increasing working-memory load required for maintaining the distinct elements that compose sequences, because this interpretation does not account for increased phasic activations observed at sequence termination. Phasic activations cannot be simply explained by switching between two successive, distinct elements composing behavioral sequences (motor responses and task sets), because phasic activations were also found in trials requiring subjects to repeat the preceding motor response or task set (sequence initiation trials). Furthermore, sustained activations are unlikely to reflect the maintenance of behavioral rules unrelated to the serial organization of motor and task sequences. Indeed, these behavioral rules, event contingencies, and performance durations were virtually identical in baseline and sequence execution periods. In both periods, subjects had to prepare the next action (movement or task sets), to respond to stimuli, to check color cues (for detecting start or stop cues), and so on. The only difference was that baseline and sequence periods involved sequences of identical versus distinct actions, respectively. This difference was associated with increased phasic activations in sequence increment compared with baseline trials. Finally, these left IPL activations might result from an increase in attentional demand during sequence execution (Cabeza and Nyberg, 2000; Cabeza et al., 2000; Corbetta et al., 2000; Culham and Kanwisher, 2001; Behrmann et al., 2004). This interpretation is unlikely, however, because attentional control engages both frontal and parietal regions (Cabeza and Nyberg, 2000), and, as mentioned above, no frontal regions showed sustained activations in both motor and task sequences.

In summary, our results provide evidence that the left middle IPS processes the abstract serial structure of behavioral plans regardless of their sensorimotor content and their hierarchical organization. This interpretation is consistent with previous results showing the involvement of the left IPS region in rehearsal of motor sequences versus rest condition (Catalan et al., 1998), of ordinal sequences versus temporal sequences (Bengtsson et al., 2004), and in encoding motor sequences at an abstract, effectors-independent level (Grafton et al., 1998). Also, left IPS activations were consistently reported during execution of structured compared with unstructured sequences (Bor et al., 2003). This interpretation is also in accordance with the previously proposed hypothesis that the left parietal region is involved in shifting the focus of motor attention from one movement in a sequence to the next (Rushworth et al., 1997).

The right homolog of the left, middle IPS region also showed sustained activations during the execution of both motor and task sequences. However, the right IPS region showed a distinct pattern of phasic activations. This region responded to task increment and motor sequence boundaries but was not sensitive to motor increments and task sequence boundaries. This activation pattern indicates that the region responded regardless of sequence boundaries and increments but only when subjects switched between simple action chunks [i.e., between either parallel sets of sensorimotor associations (task sets in the task condition) or serial sets of single acts (motor sequences in the motor condition)] (see Materials and Methods, Serial versus hierarchical organization of action sequences). Indeed, no phasic activations were observed when switching occurred only between single motor acts (motor increment) or when there was no or little task switching (task sequence boundaries included 72% trials with no task switching). Sustained activations are then likely to reflect the maintenance of selected simple action chunks in preparation of subsequent trials. In the motor condition, the entire sequence of motor acts is maintained from initiation to termination. In the task condition, a sensorimotor mapping is prepared in each increment trial and maintained up to the subsequent trial. This interpretation is consistent with a previous study showing right IPS activations associated with task-set preparation (Brass and von Cramon, 2004a). Moreover, contiguous to the right IPS region, a more ventral IPS region had the same pattern of phasic responses but exhibited sustained activations only during the execution of motor sequences. In motor compared with task sequences, motor responses could be additionally prepared between successive steps, suggesting that the right ventral IPS region is involved in further preparing single motor acts composing motor sequences. Consistently, we previously reported a similar activation pattern in the frontal regions involved in motor preparation, namely the supplementary motor area complex (Tanji and Shima, 1994; Koechlin and Jubault, 2006). Thus, our data provide evidence that the dorsal and ventral right IPL regions are engaged in preparing sensorimotor sets and motor acts, respectively, a function implicated but not specific to the execution of behavioral sequences (Asari et al., 2005; Shafritz et al., 2005). Finally, it is worth noting that we did not find any regions showing sustained activations in task sequences only, confirming that sustained activations are not segregated according to the hierarchical level of behavioral sequences.

Contiguous to these left and right IPS regions exhibiting sustained activations, additional IPS regions showed only phasic responses. These phasic responses as well as those found in the precuneus were also observed in contiguous sustained activations, except that all these regions additionally responded to motor increment. Importantly, brain regions that exhibited only phasic responses were also found in the frontal lobes (Koechlin and Jubault, 2006). More specifically, frontal regions responding to motor increment, as in these parietal regions, were found in the lateral premotor cortex only (BA 6). Posterior prefrontal regions (BA 44), in contrast, responded to motor sequence boundaries and task increment only (i.e., to boundaries of simple action chunks), whereas slightly more anterior prefrontal regions (BA 45) responded to task sequence boundaries only (i.e., boundaries of superordinate action chunks; see Materials and Methods, Serial versus hierarchical organization of action sequences). Thus, the lateral frontal cortex was proposed to form a functional system processing the hierarchical rather than serial structure of action plans (Koechlin and Jubault, 2006) with BAs 6, 44, and 45 involved in selecting single motor acts, simple and superordinate action chunks, respectively. In contrast, in parietal regions exhibiting only phasic responses, we did not find any evidence supporting such functional segregations based on the hierarchical levels of action plans, especially because all IPS regions and the precuneus responded to motor increments. In summary, this indicates that in contrast to prefrontal regions (i.e., BA 45), no parietal regions are specifically sensitive to recursive chunking of action segments (i.e., as in superordinate action chunks; see above).

An alternative hypothesis is that these IPS regions simply serve as input/output interfaces between the left IPS region, representing the serial structure of action plans, and the right IPS region, preparing successive actions (see above). This idea is supported by the fact that these IPS regions with sustained responses showed similar phasic responses. In particular, during sequence execution, the precuneus may be involved in coupling the serial-order and preparatory representations described above in the left and right IPS. Consistently, systematic connections between the precuneus and adjacent parietal areas including the IPL have been reported previously (Leichnetz, 2001), as well as the involvement of the precuneus in processing sequential complexity (Sadato et al., 1996) and in highly integrated functions irreducible to simple visuospatial processing (Cavanna and Trimble, 2006). The IPS regions exhibiting phasic responses may also serve as attentional gates between the IPL and the frontal cortex coupling serial-order and preparatory representations in the IPL with the hierarchical organization of action in the frontal cortex (Bunge, 2004). This hypothesis is supported by the differential responses to motor increments found in the posterior and anterior left IPS regions that somewhat resembles the functional segregation previously observed between premotor and prefrontal regions (see above) (Koechlin and Jubault, 2006).

Finally, our results reveal a bilateral region located in the inferior part of the SMG/TPJ that responded only to initiation and endogenous termination of both motor and task sequences. Again, in these regions there were no significant differences between motor and task sequences. Moreover, this pattern of responses is unique and was found in no other brain regions. Unlike exogenous termination, endogenous termination corresponded to stop signals occurring at the end state of learned sequences. Similarly, initiation trials corresponded to the start state of learned sequences, so that the region responded only to the “true” boundaries of memorized sequences. Thus, the SMG/TPJ region appears to be associated with the long-term memory of behavioral sequences. In the left hemisphere, this brain sector has also been found in processing linguistic sequences (Aboitiz and Garcia, 1997; Sakai et al., 2001). However, SMG/TPJ activations reported here are unlikely to reflect any covert verbal processing, because activations were bilateral and associated with sequence boundaries only. Instead, we interpret such phasic responses as reflecting retrieval of long-term representations of learned behavioral sequences when start cues occur and intrinsic inhibition of ongoing behavioral sequences when memorized end states are reached.

In conclusion, our results suggest a system of parietal regions located in the IPS and involved in the serial organization of behavior (Fig. 6). In this system, the left IPS processes the abstract serial structure of action plans that guides and keeps track of sequential progression within behavioral plans, regardless of their hierarchical structure. The right IPS is involved in successively preparing simple action chunks (dorsal IPS) and, if possible, single motor acts (ventral IPS) that compose such behavioral plans. This parietal system strongly differs from the frontal system controlling action selection with respect to the hierarchical rather than serial structure of behavioral plans (Koechlin and Jubault, 2006). Thus, our results suggest a major functional segregation within the frontoparietal network mediating action planning (Bunge et al., 2002; Rushworth et al., 2003; Brass and von Cramon, 2004b; Brass et al., 2005), with the frontal and parietal sector involved in processing the hierarchical and serial structure of action plans, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Schematic diagram of parietal and frontal activations observed in the protocol. Parietal activations are represented as solid circles and rectangles. Green and yellow colors indicate the interaction between motor increments and sequence boundaries effects in the anterior and posterior left IPS (green, larger effect of motor increment; yellow, larger effect of motor sequence boundaries). Frontal activations found in the same experimental protocol and reported by Koechlin and Jubault (2006) are shown as open circles and rectangles. Pcu, Precuneus; pSMA, presupplementary motor area.