Ventral and dorsal stream contributions to the online control of immediate and delayed grasping: A TMS approach

Abstract

According to Milner and Goodale's theory of the two visual streams, the dorsal (action) stream controls actions in real-time, whereas the ventral (perceptual) stream stores longer-term information for object identification. By this account, the dorsal stream subserves actions carried out immediately. However, when a delay is required before the response, the ventral (perceptual) stream is recruited. Indeed, a neuroimaging study from our lab has found reactivation of an area within the ventral stream, the lateral occipital (LO) cortex, at the time of action even when no visual stimulus was present. To tease apart the contribution of specific areas within the dorsal and ventral streams to the online control of grasping under immediate and delayed conditions, we used transcranial magnetic stimulation (TMS) to the anterior intraparietal sulcus (aIPS) and to LO. We show that while TMS to aIPS affected grasp under both immediate and delayed conditions, TMS to LO influenced grasp only under delayed movement conditions. The effects of TMS were restricted to early movement kinematics (i.e. within 300 ms) due to the transient nature of TMS, which was always delivered simultaneous with movement onset. We discuss the implications of our findings in relation to interactions between the dorsal and ventral streams.

Introduction

Since 1992 Milner and Goodale have proposed and refined a highly influential theory that two cortical visual streams, the ventral and dorsal stream, separately subserve perception-for-vision and perception-for-action, respectively (for reviews, see Goodale & Milner, 1992; Milner and Goodale, 1995, Milner and Goodale, 2006, Milner and Goodale, 2008). In addition, they argued that these two visual streams operate on different time scales. The ventral stream, projecting from striate cortex to infero-temporal cortex, represents an object over time, allowing object characteristics to be maintained and thereby aiding object recognition across different viewing conditions. The dorsal stream, projecting from striate cortex to posterior parietal cortex, mediates actions towards targets that are likely to continuously change, and therefore it works in real-time for immediate use in guiding actions.

Evidence to support this theory has been seen in patients with visual form agnosia and optic ataxia. Patients with damage to structures in the ventral stream may develop visual form agnosia, demonstrating profound problems with object recognition through vision. In particular, visual form agnosic patient D.F. shows movement errors if a delay is introduced between the visual stimulus and motor response (Goodale, Jakobson, & Keillor, 1994; Milner, Dijkerman, & Carey, 1999), but she shows intact performance on tasks requiring immediate action (Goodale, Meenan, et al., 1994; Rice, McIntosh, et al., 2006). Although most of the published evidence for this finding comes from a single patient, D.F., these results have now been replicated in a second patient, M.C., with impaired object recognition despite intact motion detection (Goodale et al., 2008). In contrast, damage to structures in dorsal stream regions may lead to optic ataxia, characterized by the opposite pattern of impairment; that is, they show intact performance when a delay is introduced between the stimulus and response and impoverished performance when the response immediately follows the stimulus (Goodale, Meenan, et al., 1994; Himmelbach & Karnath, 2005; Milner, Paulignan, Dijkerman, Michel, & Jeannerod, 1999; Milner et al., 2001; Milner, Dijkerman, McIntosh, Rossetti, & Pisella, 2003; Revol et al., 2003, Rice et al., 2008, Schindler et al., 2004). Importantly, aspects of this double dissociation have been demonstrated on a variety of tasks including saccades, reaching, grasping, and obstacle avoidance. It is therefore parsimonious to interpret these studies in the context of the two visual streams model, which posits that immediate motor control is mediated by the dorsal stream, yet when a delay is required this representation decays and is supplemented by a ventral stream representation.

While the research from the patient domain shows convincingly that the ventral and dorsal streams separately subserve delayed and immediate actions, respectively, it remains unclear which regions within the ventral and dorsal streams causally contribute to this behavior. In addition, while a dissociation between immediate and delayed actions has been shown in healthy subjects at the behavioral level (Hu, Eagleson, & Goodale, 1999; Hu & Goodale, 2000), it is not clear how this double dissociation is explained in the healthy brain. While the lesion data is informative in this respect, the number of patients showing the effects is small and post-lesion neuronal reorganization makes it challenging to be certain about the roles of specific brain areas in normal visuomotor performance.

Studies using neuroimaging (Binkofski et al., 1998, Culham et al., 2003; Frey, Vinton, Norlund, & Grafton, 2005), and more recently TMS (Glover, Miall, & Rushworth, 2005; Rice, Tunik, & Grafton, 2006; Rice, Tunik, Cross, & Grafton, 2007; Tunik, Frey, & Grafton, 2005) have been informative in identifying posterior parietal regions in the healthy human brain involved in grasping actions. In particular, the anterior intraparietal sulcus (aIPS) is consistently identified for its role in mediating grasping in both monkeys (Murata, Gallese, Luppino, Kaseda, & Sakata, 2000; Sakata, Taira, Murata, & Mine, 1995) and humans (Binkofski et al., 1998, Culham et al., 2003, Frey et al., 2005). Human aIPS is located at the junction between the postcentral sulcus and in the intraparietal sulcus (IPS) (for review see Tunik, Rice, & Grafton, 2007). In a series of recent studies, we have shown that transient TMS-induced disruption of aIPS (but not other parietal regions) impairs the online control of grasping with the contralateral hand (Tunik et al., 2005; Rice, Tunik, et al., 2006; Rice et al., 2007). In particular, we have shown that aIPS is causally involved in grasping when movement initiation immediately follows a ‘go’ cue or when an immediate movement correction to a perturbation is required (Rice, Tunik, et al., 2006). We posited (Rice, Tunik, et al., 2006) that the role of aIPS in the online control of grasping may be related to the computation of a difference vector, which is computed by determining the difference between the target vector and current position vector (Ulloa & Bullock, 2003).

Although behavioral and neuropsychological evidence suggests that delayed, but not immediate, actions recruit the ventral stream, it remains uncertain which ventral stream regions are critical. One likely candidate is the lateral occipital complex, LOC (Malach et al., 1995), which includes the lateral occipital area, LO, on the lateral surface of the occipito–temporal junction, along with other areas on the ventral occipitotemporal (VOT) surface (such as the posterior fusiform sulcus). The LOC is activated in human neuroimaging studies across a range of object perception and recognition tasks (for review see Grill-Spector, Kourtzi, & Kanwisher, 2001). Moreover, TMS to LO (Brodmann's area 37) slows subjects’ reaction times for object picture naming (Stewart, Meyer, Frith, & Rothwell, 2001) and shape discrimination (Ellison & Cowey, 2006). Interestingly, LO is damaged bilaterally in D.F. (James, Culham, Humphrey, Milner, & Goodale, 2003), a patient with visual form agnosia who can perform visually guided grasping in real time, but not following a delay (Goodale, Jakobson, et al., 1994). D.F.’s case suggests that although immediate grasps can be performed by the dorsal stream (including aIPS, which remains intact) in the absence of LO input, delayed grasps appear to require that LO be intact. Note that D.F.’s lesions do not include all of LOC; the lateral subdivision (LO) is damaged, but the VOT subdivision is spared (James et al., 2003; Steeves, Humphrey, Culham, Menon, & Goodale, 2002; Steeves et al., 2004). Another possibility, is that some of D.F.’s deficits arise from an additional lesion in left medial parieto-occiptial cortex rather than LO (James et al., 2003). Patient M.C., who also shows impairments for delayed but not immediate grasping (Goodale et al., 2008) also has damage to LO; however, like D.F. she also has additional lesions (in this case, bilateral occipital and right parietal cortex) (Culham, Witt, Valyear, Dutton, & Goodale, 2008), making it difficult to definitively show that LO is the critical area accounting for dissociation between immediate and delayed grasping.

Evidence from functional MRI supports the suggestion that LO is involved in delayed but not immediate grasping. First, fMRI during immediate actions has indicated that aIPS is reliably activated for visually guided grasping (in which vision is used to preshape the hand) compared to reaching (in which no preshaping is required); whereas, LO is equally activated by grasping and reaching, suggesting that real-time grasping requires processing in aIPS but not LO (Cavina-Pratesi, Goodale, & Culham, 2007; Culham et al., 2003). Second, fMRI during delayed grasping has revealed reactivation of LO and early visual areas at the time of action execution, many seconds after the object was visually presented (Singhal, Kaufman, Valyear, & Culham, 2006), suggesting that LO is involved in online control of an action under delayed conditions. Activation in LO and early visual areas was no higher during the delay period than during the intertrial interval. These results suggest that while aIPS may suffice to control grasping in real-time, after a delay, information about object shape, size and orientation may be recruited from LO and early visual areas to control the grasp. If this hypothesis is true, then immediate grasping should be unaffected by TMS to LO while delayed grasping should be disrupted if the stimulation occurs once the delayed action is cued.

However, at least three other outcomes are possible. One possibility is that the reactivation of LO after a delay is merely “epiphenomenal”, perhaps occurring because subjects are imagining the objects to be grasped even if visual imagery is not essential to the success of delayed grasping. Another possibility is that delayed grasping is mediated by aIPS rather than LO. Macaque neurophysiology within area AIP (the proposed functional equivalent of human aIPS) has reported sustained activation during the delay period (Murata, Gallese, Kaseda, & Sakata, 1996), a result corroborated by human fMRI (Singhal et al., 2006). This would lead to the alternative hypothesis that TMS to aIPS would disrupt both delayed and immediate grasping but TMS to LO would disrupt neither. A third possibility is that other areas within the ventral stream besides LO are recruited by delayed grasping. Delayed grasping, in which the target and action are offset in time, may be similar in spirit to pantomimed grasping, in which the target and action are offset in space (i.e. subjects pretend to perform a grasp in a location adjacent to the object while the object remains present (Goodale, Jakobson, et al., 1994). An fMRI study of pantomimed vs. real actions found activation in another region of the ventral stream – the middle temporal gyrus – rather than LO (Króliczak, Cavina-Pratesi, Goodman, & Culham, 2007) In sum, although there is evidence to suggest that delayed grasping requires revival of object-specific information from LO, a new experiment with TMS offers the possibility to test the various hypotheses, but to make stronger arguments (than fMRI) based on the necessity of aIPS and LO to immediate and delayed actions.

Given the two streams theory, we tested the hypothesis that processing within aIPS and LO may be particularly important in the online control of immediate versus delayed grasping, respectively. We applied TMS to induce a “virtual lesion” to aIPS or LO at the time when subjects executed grasping movements to targets of varying size. We chose to induce stimulation at the time of action based on our prior neuroimaging results suggesting that LO is reactivated at the time of action, but not differentially activated during the delay period (Singhal et al., 2006), and because our prior TMS data has shown that aIPS becomes necessary for grasping at the time of movement execution (Rice, Tunik, et al., 2006), reflecting a specific role in the online control of the movement. The auditory cue to initiate the grasping movement either immediately followed vision of the target or occurred after a two second delay. Grasping was performed in open loop, using liquid crystal shutter glasses to control viewing time of the target. We choose this behavioral paradigm, for several reasons. First, our previous TMS investigations have confirmed that the dorsal stream is necessary for immediate grasping under these behavioral conditions (Rice, Tunik, et al., 2006; Rice et al., 2007). Second, a recent study has shown that an optic ataxia patient (with unilateral dorsal stream damage) is unable to avoid obstacles in his workspace when the movement is performed in open loop, simultaneous with visual occlusion; yet this behavior improves following a delayed response (Rice et al., 2008). Finally, we believe that this paradigm has advantages over some previous paradigms investigating delayed actions, which have involved contrasting: (1) immediate closed loop motor conditions to delayed open loop motor conditions (Milner, Paulignan, et al., 1999); (2) immediate real grasping to delayed pantomimed grasping (Goodale, Jakobson, et al., 1994; Milner et al., 2003). In our study, both the viewing time of the target and the vision of the hand and arm were identical in our immediate and delayed conditions, meaning that any differences observed could be not be accounted for by confounding influences, but instead would be related to how the movement representation (current in the case of immediate grasping; and a perceptual memory representation in the case of delayed grasping) influences the online control of the movement.

We predicted, based on the two streams model and the fMRI reactivation of LO upon action execution, that relative to the no-TMS condition, TMS-to-aIPS would interfere with grasping under immediate conditions only, whereas TMS-to-LO would interfere with grasping under delayed conditions only. Given that past studies have shown that TMS to aIPS affects the kinematic parameters of the grasp but not the transport (Rice, Tunik, et al., 2006; Rice et al., 2007; Tunik et al., 2005) we were particularly interested in the grasp-related dependent variables (maximum grip aperture and the time at which it occurred; peak velocity and the time at which it occurred). By looking at both these variables we were able to assess the TMS effects on early movement kinematics (peak velocity) and late movement kinematics (maximum grip aperture). This is particularly important as the effects of TMS are known to be transient and as such would not result in the gross behavioral deficits observed in patient studies.