Dependence of Manual Grasping on the Behavioral Context : A Comparison between Arms and between Age Groups

Copyright © 2013 Otmar Bock, Benjamin Baak. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. In accordance of the Creative Commons Attribution License all Copyrights © 2013 are reserved for SCIRP and the owner of the intellectual property Otmar Bock, Benjamin Baak. All Copyright © 2013 are guarded by law and by SCIRP as a guardian.


Introduction
It has been suggested before that findings on motor performance, yielded in the laboratory, may not necessarily apply in everyday life (Chaytor & Schmitter-Edgecombe, 2003;Ingram & Wolpert, 2011).We have recently scrutinized this view by asking subjects to grasp and move a lever either in a typical laboratory context (L)-grasping was instructed, externally triggered, repetitive, and served no ultimate purpose-or in a more everyday-like context (E)-grasping was not explicitly instructed, self-initiated, embedded in complex behavior and had the ecologically valid purpose to earn money (Bock & Hagemann, 2010).Even though the mechanical constraints were identical in both contexts, movement kinematics and dynamics widely differed.These differences could not be reduced to a single underlying cause since factor analysis yielded multiple orthogonal factors (Bock & Züll, 2013).This led us to conclude that grasping is controlled by multiple functional modules which are differently sensitive to context.
Further research revealed that context-sensitivity can't be reduced to differences between L and E regarding movement speed, attention focusing or task complexity, since manipulations of those differences didn't consistently change context-sensitivity (Steinberg & Bock, 2013c).However, we observed a consistent effect of personality traits: context-sensitivity was accentuated in subjects who prefer slow, attentive and prudent processing (Steinberg & Bock, 2013b).We proposed that this processing style is characteristic for the ventral rather than the dorsal occipito-frontal stream in the human cortex (Goodale & Milner, 1992;Milner & Goodale, 1993).while the dorsal stream engages in quick automated reactions, the ventral stream specializes in slow and attention-demanding behavior (Buxbaum, Johnson-Frey, & Bartlett-Williams, 2005;Daprati & Sirigu, 2006;Rossetti & Pisella, 2002).Our interpretation thus links context-sensitivity to different cortical processing streams.
The present study investigates whether context-sensitivity, observed previously for the dominant arm, holds equally for the non-dominant arm.It is well established that movement performance is not the same for both arms: the non-dominant arm controls intersegmental torques less well (Sainburg & Kalakanis, 2000), which shows poorer performance on tasks requiring high precision (Gonzalez, Ganel, & Goodale, 2006;Wing, Turton, & Fraser, 1986;Woodworth, 1899), but better performance than the dominant arm on tasks requiring high speed (Annett, Annett, Hudson, & Turner, 1979;Carson, Chua, Goodman, Byblow, & Elliott, 1995;Elliott et al., 1993).It has been concluded that the non-dominant arm is optimized for controlling limb posture, and the dominant arm for regulating limb trajectory (Sainburg, 2004).Such a specialization might reflect the preferred arm use in bimanual activities: objects are typically held and stabilized by the non-dominant, and manipulated by the dominant hand (Grosskopf & Kuhtz-Buschbeck, 2006;Trevarthen, 2010).Arm specialization has also been linked to differences in the underlying control principles, feedback control for the non-dominant versus preplanning for the dominant arm (Sainburg & Kalakanis, 2000).Given these profound differences between arms, we reasoned that context-sensitivity might also be different.Specifically, we formulated two alternative hypotheses: according to the first, the non-dominant arm depends more heavily on sensory feedback and therefore should be less susceptible to extraneous influences such as behavioral context.
According to the second, the non-dominant arm is not wellpracticed in manipulation tasks such as grasping and therefore should be more susceptible to extraneous influences such as behavioral context.
We have shown before that context-sensitivity of grasping changes in old age; some parameters become more and other less context-sensitive, with no substantiable net change across all parameters (Bock & Steinberg, 2012).Again, these data have been yielded in the dominant arm.Since handedness is less pronounced in old age (Kalisch et al., 2006), we expected that any increase or decrease of context-sensitivity observed in the left arm of young subjects should be smaller in the left arm of seniors.To find out, the present study includes data from the non-dominant arm of elderly participants.

Participants
Forty-eight young (24.3 ± 3.9 years) and thirty older subjects (71.8 ± 7.4 years) participated.All were right-handed, free of musculoskeletal impairments, diseases of the nervous system and visual deficits except for corrected vision by self-report, and lived independently in the community.None of them had participated in research on grasping or cognition within the last 12 months.An ethical approval for this study was given by the institutional review board of the German Sport University Cologne, and all subjects signed an informed consent statement before participating.Half of the young subjects were tested using their dominant (right) arm.The other half of the young and all older subjects were tested using their non-dominant (left) arm.

Task and Procedure
Experimental hardware and procedures were as in our previous studies (Bock & Beurskens, 2010).Subjects sat at a table facing a 17 computer screen 67 cm ahead.A cylindrical lever of 4 cm length and 1.5 cm diameter was positioned 35 cm away from the front edge and 16 cm above the surface of the table 10 cm to the right of the screen or, for left arm testing, 10 cm to the left.The lever was covered by a hood from three sides, to ensure that subjects could only grasp it with the precision grip (thumb and index finger).The lever could slide 3.5 cm towards the subjects' body midline along a rail (see Figure 1), where it met a mechanical stop.A displacement sensor (Burster ® 8740) registered the lever's position and a 6 df force transducer (ATI ® Nano 17) registered the forces applied to the lever, both with a sampling rate of 250 Hz.A joystick was mounted 41 cm in front of the screen with its tip 12 cm above the table's surface, such that its distance from the lever was 32 cm horizontally and 4 cm vertically.Six reflecting markers of 6 mm diameter were placed on thumb and index finger of the subjects' grasping hand with double-sided adhesive tape, and two Vicon ® MX-F20 3D high resolutions infrared cameras (sampling rate: 250 Hz, 1680 × 1280 pixels) registered their positions.
In a laboratory task (L), the joystick was locked in its central position and subjects touched its tip with thumb and index finger.At randomly varying intervals of 2 -6 s, a green dot was displayed on the screen accompanied by a beep, prompting subjects to release the joystick and grasp the lever, to slide it towards them and back again, and then to return the hand to the joystick.In an everyday-like task (E), the joystick was unlocked and subjects were asked to play a computer game of chasing spiders on the screen with a joystick-driven cursor.A reward of .02€ was displayed near the right edge of the screen for each spider hit.Each game level terminated after 10 s, and subjects then had to collect their reward by moving the joystick to the center, grasping the lever, moving it towards them and back, and then returning their hand to the joystick.To keep the game motivating, speed and complexity of spider movement increased after every fifth level.No instructions were given on how to grasp the lever in task E, and subjects were not told that the purpose of their participation was to collect data on grasping.In accordance with our earlier study (Bock & Steinberg, 2012), spider speed was 30% lower in older subjects than in young ones.
To exclude carryover effects, each subject was engaged in only one of the tasks.Thus, 12 young subjects were tested in L and 12 in E using their left arm, 12 young subjects were tested in L and 12 in E using their right arm, and 15 older subjects were tested in L and 15 in E using their left arm.Each subject had 3 -5 practice trials with the pertinent task and hand, to ensure that procedures were understood, and data were then collected for 20 grasping responses per subject.Note that both tasks used the same objects (joystick and lever) in the same location, and required the same hand and lever movements; they only differed with respect to their context: grasping was instructed, repetitive and served no ultimate purpose in L, but was uninstructed, part of complex behavior and had financial gain as purpose in E.

Data Analysis
Registered data were reduced by an interactive computer algorithm to 20 parameters representing the means of kinematic and dynamic landmarks across trials, and 20 parameters representing the pertinent coefficients of variation.The additional parameter "Peaks" can't be parsed into a mean and a CV.A definition of all parameters is provided in Table 1.Each parameter was submitted to a two-way analysis of variance (ANOVA).Left and right arm performance of young subjects was compared with the between-factors Arm (left, right) and where l rh , e rh , l lh and e lh are parameter values of individual sub-jects participating with their right hand in task L or E, or with their left hand in task L or E, respectively, while L RH , E RH , L LH and E LH are the corresponding group means.Thus, large  scores represent a stronger task-dependence of the right compared to the left hand.Significant effects of Age * Task were transformed accordingly.We then submitted the scores to factor analyses with varimax rotation, using the inclusion criterion F = 1.

Results
The right part of Table 1 summarizes the ANOVA outcome for young subjects using the left or right arm in L or E.
As in our previous work (Bock & Steinberg, 2012;Bock & Züll, 2013;Steinberg & Bock, 2013a;Steinberg & Bock, 2013b), the effect of Task was significant for a number of parameters.The effect of Arm was significant for several parameters as well, and that of Task * Arm was significant for nine parameters.The latter were transformed into scores, and were then reduced by factor analysis to three orthogonal factors, explaining 64.5% of total variance (see Table 2).
To obtain a global measure of Task * Arm effects, we normalized each parameter p with Task * Arm significance by and then calculated the rms value of p' across parameters.
The outcome is depicted in Figure 2(a): the nine parameters with Task * Arm significance were task-independent for the right, but task-dependent for the left hand.
We then replicated the same procedure for factor rather than parameter values, yielding the outcomes in Figures 2(b)-(d): the same pattern described above also emerged for each factor.
The right part of Table 1 summarizes the ANOVA outcome for young and older subjects using the left arm.Again, a number of parameters showed significant effects of Task and/or Age, and eight parameters yielded significant effects of Task * Age.The latter were reduced by factor analysis to three orthogonal factors explaining 69.29% of total variance (see Table 3).The right part of Figure 2 illustrates that parameters with significant Task * Age interactions were task-dependent in young but not in older subjects, and that this is reflected by all three factors, although to a varying degree.

Discussion
The purpose of our study was to evaluate the role of context when grasping with the non-dominant arm.The outcome confirms once more that the kinematics and dynamics of grasping are context-sensitive (effects of Task in Table 1), and documents differences for grasping with the left versus right arm (effects of Arm in Table 1).The arms differed with respect to speed and accuracy, as expected from literature (see Introduction), but they also differed with respect to path shape (detour-H and detour-V) and final hand posture (incli-end).The latter findings can't be explained by biomechanical constraints since the task was exactly mirror-symmetrical for the two arms, and rather support the existence of different control principles for the two arms (Grosskopf & Kuhtz-Buschbeck, 2006;Sainburg & Kalakanis, 2000;Trevarthen, 2010).Most importantly for the purposes of our study, context-sensitivity was not the same for both arms (effects of Task * Arm in Table 1).Specifically, nine parameters showing no context-sensitivity for the right arm did show such sensitivity for the left arm; this result emerged when all nine parameters were considered together, and also when each grasping factor was considered separately.From this we conclude that context played a larger role for the left arm, as stipulated by our second hypothesis, and not a smaller role, as stipulated by the first hypothesis (see Introduction).We therefore discard the view that the non-dominant arm is less influenced by context because of its stronger reliance on sensory feedback, and rather adopt the alternative view that it is more influenced by context because of its low experience with manipulation tasks such as grasping.Obviously, further work will be needed to substantiate this view.One possible approach could be to compare right-and left-handed subjects in our paradigm.Since righthanders strongly prefer to grasp with their right arm while lefthanders exhibit no arm preference (Gon-   zalez, Whitwell, Morrissey, Ganel, & Goodale, 2007a), the nondominant arm of lefthanders is experienced with manipulation tasks and their Task * Arm effects should therefore be less pronounced than in righthanders, if our second hypothesis is indeed correct.
In our previous work, we have related context-sensitivity to the existence of two occipito-frontal processing streams in the human cortex: A dorsal stream is mainly concerned with fast automated reactions, and a ventral stream dealing with slow, attention-demanding behavior (Buxbaum et al., 2005;Goodale & Milner, 1992;Rossetti & Pisella, 2002).Since both streams are interconnected (Goodale & Westwood, 2004), processing of a given sensorimotor action may not be exclusively confined to one of the two streams, but may involve both of them in varying degrees.Given these facts, we posit that young subjects using their right arm in L will preferentially engage the dorsal stream, since L requires externally triggered, stereotyped behavior.In E, however, they will more strongly involve the ventral stream since E requires complex, volitional behavior.This view is illustrated in a simplified fashion by the top half of Figure 3.The bottom half of that figure illustrates how the neural activation might change when young subjects use their left arm.Since arm is specialized for postural rather than volitional responses (see Introduction), its control circuitry might be well suited for the automated responses in L, but might require particularly strong ventral activation for the volitional responses in E. Accordingly, Figure 3 shows no difference between left and right arm in L, but a shift towards the ventral stream for the left arm in E. Note that as a consequence, context sensitivity (i.e., the difference between L and E) is more pronounced for the left than for the right arm, as observed experimentally.
Our data further show that grasping performance is affected by old age (effects of Age in Table 1).Elderly persons expectedly differ from younger ones regarding movement duration and variability, but also regarding path shape and hand posture.Most importantly for the purposes of our study, context-sensitivity of the left arm differed between age groups (effects of Task * Age in Table 1): eight parameters showing context differences in young subjects showed smaller, null or even inversed context differences in the elderly, with the net effect across all parameters being an absence of an appreciable context-sensitivity.In other words, the increase of context-speci- ficity from the dominant to the non-dominant arm, as observed in young subjects, was attenuated if not absent in the elderly.This conforms to our expectation (see Introduction), according to which less pronounced handedness in old age is paralleled by less pronounced differences between the two arms regarding context-sensitivity.Referring back to Figure 3, one could argue that seniors grasping in E can't increase the ventral contribution when using their left rather than their right arm, and their performance with the left arm therefore resembles that of young subjects using their right arm.

Figure 2 .
Figure 2.Root mean square values across parameters with significant Task * Arm effects (a, a') and for each constituent factor(b-d, b'-d').Graphs at the left illustrate the differences between right and left hand in young subjects, and those to the right the differences between the left hand of young and older subjects.Symbols represent averages across subjects, and error bars the pertinent interindividual standard deviations.
2 & 3, Numbers are factor loadings, only values ≥ 0.6 are shown.The bottom row indicates the fraction of total variance explained by the respective factor.GF stands for grasping factor.

Figure 3 .
Figure 3. Schematic representation of assumed sensorimotor processing through the dorsal and the ventral stream in task L and E, when young subjects use their right versus left arm.

Table 1 .
Parameter definitions and the pertinent ANOVA outcomes * ., ** and *** Represent p < 0.05, p < 0.01 and p < 0.001, respectively.For each parameter, ANOVA for the means is presented above ANOVA for the CVs.
* Numbers are F-values with 1.44 degrees of freedom.*

Table 2 .
Outcome of factor analysis for parameters with significant Task * Arm effects * .

Table 3 .
Outcome of factor analysis for parameters with significant Task * Age effects * .