Two types of control mechanisms are used to perform skilled hand movements: predictive mechanisms – that enable one to anticipate the movement on the basis of sensorimotor memories of the manipulated object, and reactive mechanisms – that allow for correction of the movements/forces. Reactive mechanisms are typically feedback loops emanating from sensory and/or visual afferents. It is generally accepted that predictive mechanisms, also called feedforward or anticipatory mechanisms, are based on internal models in the central nervous system (Wolpert and Ghahramani, 2000). These include: (1) an inverse model that uses the current state of the limb and the specific context to generate an appropriate motor command, (2) a forward sensory and motor model that predicts the movement resulting from the motor command and estimates the sensory feedback of the new state of the limb, allowing a comparison (3) with actual feedback (4) and subsequent adjustment of the motor command (Figure 1). It is the close interaction between predictive and reactive mechanisms that allows for the production of smooth movements.

In the context of brain lesions, prehension deficits could be explained by impairments in predictive and/or reactive mechanisms. This is of particular interest for the design of rehabilitation programs. These mechanisms have been well-studied and described in healthy adults and typically developing children, especially through analysis of the timing and magnitude of grip force (GF) and the tangential load forces (LF) during grip–lift manipulation (Johansson and Westling, 1984, 1988a; Forssberg et al., 1991, 1992, 1995; Gordon et al., 1992; Witney et al., 2004). In healthy adults, there is a well-described sequence of events comprising a typical coordination pattern (Johansson and Westling, 1984, 1988a) during a grip–lift task (see Figure 2): first the contact between fingers and the object is initiated in a quick succession (T0–T2). During the preload phase (T2–T3), GF increases prior to LF onset. GF and LF subsequently increase in parallel during the loading phase (T3–T4). Forces rates are characterized by single peaks that are well-timed. After the parallel increase of forces, the forces are steadily maintained during a static phase (T5–T6) in which the object is held in the air, that is followed by the release of the object including a replacement phase (T6–T7) an a subsequent rapid decrease in the grip and LF (T7–T8) until the thumb and index fingers are released from the object (T8–T9).

The adaptation of the forces to the object’s properties during precision grip tasks requires the use of tactile (Westling and Johansson, 1987) and weight-related information (Johansson, 1996). Adaptation to various load and frictional conditions has been well-established. However, such information about the object’s physical properties is not instantly available due to delays in the transmission of sensory information. Therefore, to manage objects with adequate forces in everyday life, GF and LF forces are planned in an anticipatory manner on the basis of internal representations of the objects from prior manipulatory experience (Johansson and Westling, 1987, 1988a; Gordon et al., 1993), i.e., before initiating the movement.

In typically developing children, this anticipatory coordination of GF and LF matures until approximating that of adults at the age of 6–8 years (Forssberg et al., 1991). Before that age, the forces do not increase in parallel, and there are negative LFs at the start and excessive and multiple increments in GF during the loading phase (Forssberg et al., 1991).

This immature strategy has been interpreted as a control strategy relying mainly on feedback. Due to the delay of the feedback, the adjustment of the motor command is sequential, which does not allow smooth movements. With continuing maturation and experience during childhood, children utilize internal representations of the objects, allowing a predictive (anticipatory) control of the movement (Forssberg et al., 1991).

The integration of the precise GF and LF coupling in a bimanual context have been studied recently when opposite or concomitant patterns of force are required in the hands: a handheld object had to be placed on the top of another object statically held in the other hand (Islam et al., 2011) or two pieces of a handheld object had to be pulled apart (Smits-Engelsman et al., 2011). These tasks, when performed with healthy adults, demonstrated coordinated actions of both hands.

The use of one hand to generate a rapid increase of LF to an object held by the opposite hand has also been used as a model of coordination. In that context, a predictive increase in GF is observed before the rapid LF increase (Eliasson et al., 1995; Bleyenheuft et al., 2009). After the brisk LF increase, there is a systematic second increase in GF, programed as a predictive action that allows an optimal stabilization of the object around the brisk force increase. Self-induced brisk load increases also mature during childhood, approaching adult values at 9–10 years of age (Eliasson et al., 1995; Bleyenheuft and Thonnard, 2010a).

PubMed and Scopus electronic databases were searched using combinations of the following terms: GF, fingertip force, and precision grip either with “cerebral palsy” or “stroke.” Studies from 1984 to 2013 were retained. An additional hand-search was conducted in the reference lists of the articles meeting the search criteria. The search procedure included only studies in English. Studies focused on heat and pain, as well as animal studies and studies dedicated to other clinical forms of CP (e.g., diplegic or bilateral CP) were excluded.

In children with HCP, the fingertip forces deviate from that observed in age-matched typically developing children: the timing of GF and LF onset is disrupted, GF is excessive, especially at the onset of LF increase (Eliasson et al., 1991, 1992, 1995a; Forssberg et al., 1999; Gordon and Duff, 1999; Gordon et al., 1999; Hermsdörfer et al., 2003) and present multiple successive increments that can also be observed in the force rates. The object is frequently pushed down toward the support, generating a negative LF before initiating the lifting phase. While encountering these impairments in precision grip, children with HCP demonstrated some ability to adapt to various weight and friction conditions, provided they had a greater number of trials to adapt to new conditions (Steenbergen et al., 1998; Gordon and Duff, 1999). As shown in a longitudinal study, these impairments are not increasing with time (Eliasson et al., 2006), which is congruent with other studies on deficits in maximal grip strength and isometric finger forces (Smits-Engelsman et al., 2004). During the release of a handheld object, children with HCP demonstrate also a sequential coordination of forces (Eliasson and Gordon, 2000). However these forces are adapted to the weight of the object (Eliasson and Gordon, 2000). Impairments in replacement and release of handheld objects are increased when speed and/or accuracy are imposed (Gordon et al., 2003). However, it is not clear if the impairments observed are linked to predictive or reactive control of precision grip.

Predictive abilities in grasp control were demonstrated in children with cerebral palsy, since they are able to form and retain internal representations of new objects (Duff and Gordon, 2003); there is an influence of bimanual tasks on the performance of the more and less-affected hand (Steenbergen et al., 2008) and these children have the ability to transfer learned forces from the less-affected to the more affected hand (Gordon et al., 1999, 2006). However these predictive abilities (Mutsaarts et al., 2006; Steenbergen and Gordon, 2006; Bleyenheuft and Thonnard, 2010b), as well as their motor imagery (van Elk et al., 2010), are clearly impaired. It has been suggested that the impaired precision grip of children with HCP is the consequence of an inability to use internal models of manipulated objects. This could be due to inefficient feedback from the paretic hand (Gordon and Duff, 1999; Gordon et al., 1999) or to high-level deficits in sensorimotor integration (Eliasson et al., 1992; Bleyenheuft and Thonnard, 2010b). The role of predictive and/or reactive deficits in the precision grip of children with HCP has been further studied using a paradigm (self-or-examiner-imposed load to a handheld object) where both were tested independently (Bleyenheuft and Thonnard, 2010b). Predictive and reactive conditions were observed in the paretic hand: in the paretic hand, children were able to anticipate events prior to the brisk load increase, but were unable to pursue predictive control afterward (Bleyenheuft and Thonnard, 2010b). Interestingly, these predictive motor control deficits were not observed in the less-affected hand. The less-affected hand had been previously studied in paradigms showing that while anticipatory control is present (Gordon et al., 1999, 2006; Steenbergen et al., 2008), subtle deficits can be observed (Steenbergen et al., 1998; Forssberg et al., 1999; Gordon et al., 1999; Mutsaarts et al., 2006; van Elk et al., 2010). These subtle deficits are congruent with those observed on the more affected hand but are not as pronounced: compared to controls, manipulation with the less-affected hand demonstrates lower force rates, increased duration of preload phase, negative LF during the preload phase, overall higher GF, and an increase of forces that is less parallel (Gordon et al., 1999). Nevertheless, anticipatory control is present (Gordon et al., 1999, 2006; Steenbergen et al., 2008) suggesting lateralized impairments in motor planning underlying precision grip in HCP, which could reflect a lateralized deficit in the sensorimotor integration (Prabhu et al., 2011). In bimanual tasks, however, children with HCP demonstrate less accurate performance in both hands when compared to unimanual use (Islam et al., 2011; Smits-Engelsman et al., 2011), suggesting that while dissociated in unimanual motor planning, each effector can be influenced by the other in a bimanual context.

These results, consistent with previous findings, suggest that the abilities of the less-affected side could be used in neurorehabilitation to improve motor control of the paretic side. For a recent systematic review on sensorimotor deficits in children with HCP, see also Bleyenheuft and Gordon (2013).

Precision grip in the paretic hand of stroke patients is also impaired. The GF is generally higher – which is also seen already at the onset of the LF increase – and prolongations are observed in the timing of the movement: there is an asynchronous onset of GF and LF, leading to a longer preload phase and discontinuous force increase (Hermsdörfer et al., 2003; Nowak et al., 2003; Takahashi and Reinkensmeyer, 2003; Wenzelburger et al., 2005; Blennerhassett et al., 2006, 2008; McDonnell et al., 2006; Raghavan et al., 2006; Quaney et al., 2010). This longer preload phase is also correlated to clinical measures of handgrip limitations (Blennerhassett et al., 2008). Similar to congenital hemiparesis, the stroke patients push the object on the table before lifting it (McDonnell et al., 2006). There is a reduced ability to adapt to changes in the object’s weight (McDonnell et al., 2006; Raghavan et al., 2006), although some adaptations still occur. The release phase is also impaired, showing sequential force coordination (Naik et al., 2011). Altered digit GF direction during prehension tasks in stroke patients has a direct impact on grip (Seo et al., 2010), but can be improved through visual force feedback (Seo et al., 2011). Deficits in prediction of the inertial load profile arising from voluntary movement with handheld objects were also demonstrated in patients with acute stroke performing lift-and-hold tasks and point-to-point movements (Nowak et al., 2003). Since the same type of impairments has been delineated in patients with cortical and subcortical stroke – increased GF and deficits in temporal coupling – the authors suggested that the internal models responsible for precise regulation of forces were impaired. While these models are thought to be formed in the cerebellum (Wolpert et al., 1998; Wolpert and Flanagan, 2001), Nowak et al. (2003) argued that cortical and subcortical structures should be involved in the subsequent process of issuing motor commands. Studies investigating the ability of patients with chronic stroke to learn anticipation (either by force disturbance during a reaching task (Wenzelburger et al., 2005), or presenting objects with different shapes (Raghavan et al., 2010) showed an impaired ability to adapt. However, a partial ability to form and use internal models was retained (Wenzelburger et al., 2005). In contrast, Hermsdörfer et al. (2003, 2004) suggested that internal models are preserved in patients with chronic cerebral stroke since the feedforward nature of the GF coupling with LF was preserved during object oscillations. The difference observed in these two paradigms may be linked to the fact that the GF–LF coordination was studied, respectively, in discrete and rhythmic coordination patterns. We hypothesize that rhythmic (more robust) coordination might be preserved while anticipatory motor control needed for discrete tasks might be impaired. These high-order motor planning deficits observed in patients with subcortical subacute and chronic stroke are ameliorated by transfer of information from the unaffected hemisphere (at least in right hemiparesis) (Raghavan et al., 2006).

The precision grip impairments described for various stroke patients are of importance because there is an impact of unimanual dysfunction on bimanual tasks, and therefore on activities of daily living (Müller and Dichgans, 1994). Excessive GF and deficits in the temporal coupling of the forces have been observed in the paretic hand of stroke patients (cortical, subcortical, and cerebellar lesions). On the basis of one of the largest studies of precision grip control after stroke, the excessive GF was attributed to impaired sensory feedback (Blennerhassett et al., 2007). The lack of temporal coupling was linked to deficits in developing internal models. Temporal coupling impairments could also be due to disturbed feedback transmission – this implies that an inability to update the sensorimotor memory. Predictive and reactive nature of deficits in precision grip have also been assessed further using the self or examiner brisk loading bimanual system described for children with HCP in order to test both of them separately. Deficits highlighted in the paretic hand were similar to those observed on the more affected hand of children with HCP. In predictive conditions, while an anticipatory control was present, temporal deficits were shown after brisk load increase. In reactive conditions, a slight increase was observed in the reflex latency. The less-affected hand did not differ from age- and sex-matched healthy controls (Dispa et al., 2014).

Previous studies including analysis of precision grip in the less-affected hand in patients with unilateral middle cerebral artery stroke (Quaney et al., 2005) showed an increase GF at lift-off across dynamic and static portions of the grip–lift task. In patients with subcortical lesion (subacute phase), the ipsilesional hand was also affected with a slowing of movements and poor coordination of grip and lift forces (Nowak et al., 2007).

Both in stroke patients and children with HCP, major precision grip impairments are observed in the paretic hand. Many of the deficits highlighted in the prehension of children with HCP and stroke patients present similarities: excessive GF, lack of GF and LF coupling, negative LF at the onset of the movement, release impairments, and a decreased ability to adapt to different loads (Table 1). These impairments are likely due to impairments in both predictive and reactive process.