Predictive force programming in the grip-lift task: The role of memory links between arbitrary cues and object weight

Abstract

We tested the ability of healthy participants to learn an association between arbitrary sensory cues and the weight of an object to be lifted using a precision grip between the index finger and thumb. Right-handed participants performed a series of grip-lift tasks with each hand. In a first experiment, participants lifted two objects of equal visual appearance which unexpectedly and randomly changed their weight. In two subsequent experiments, the change in object weight was indicated by cues, which were presented (i) visually or (ii) auditorily. When no cue about the weight of the object to be lifted was presented, participants programmed grip force according to the most recent lift, regardless of the hand used. In contrast, participants were able to rapidly establish an association between a particular sensory cue with a given weight and scaled grip force precisely to the actual weight thereafter, regardless of the hand used or the sensory modality of the cue. We discuss our data within the theoretical concept of internal models.

Introduction

Everyday we handle hundreds of objects in the environment that impose novel mechanical constraints, such as weight, shape and surface friction. When grasping and lifting an object using a precision grip between the index finger and thumb, we have to generate a grip force (exerted against the object's surfaces) which has to be sufficient to compensate for the vertical lift force (tangential to the grip surfaces) in order to prevent object slippage. On the other hand, we also have to avoid excessive grip forces as, for example, fragile objects may crush. One crucial issue then is how we adjust our grip forces in an environment offering a huge diversity of objects with various mechanical characteristics? To date, the details of the underlying behavioural processes remain to be explored. Based on our life-long experience with objects of different sizes and shapes we probably use transformational processes when lifting a novel object. After a visual analysis of the object, with emphasis on its size and shape, we use visuomotor transformations to select the necessary grip and lift forces based on a default value for object density (Gordon, Forssberg, Johansson, & Westling, 1991; Mon-Williams & Murray, 2000). Somatosensory input from the grasping fingers allows a more accurate scaling of grip and lift forces to the mechanical object properties. When lifting a given object repetitively, accurate force adjustment is typically achieved within 2–3 lifts (Johansson & Westling, 1984). Once established the memory between a given object and the grip force necessary to lift it is easily transferred across hands, suggesting generalisation from one hemisphere to the other, and also maintains in memory for up to 24 h (Gordon, Forssberg, & Iwasaki, 1994).

However, in daily life we often lift objects whose visual appearance does not allow for a prediction of its weight to select the appropriate grip forces in advance. For example, grasping and lifting a stein of Bavarian beer does not allow for a visual judgement about its content. It appears that our motor system chooses a simple way out of the dilemma: we scale our grip force to match the mechanical properties of the object we lifted most recently (Johansson & Westling, 1984; Nowak & Hermsdörfer, 2003). Such behaviour, however, results in prediction errors, i.e., we apply forces that are too low when we lift a heavier than expected object, and forces that are too high when we lift a lighter object than expected. Such predictive scaling of grip force is initiated well before lift-off until somatosensory information regarding the object's mechanical properties, i.e., its weight, is available (Flanagan & Johansson, 2002; Johansson & Westling, 1988). During lifting the object successfully, we are able to obtain the relevant somatosensory information to rapidly establish an internal representation of the mechanical object properties to be remembered and adjust grip forces accordingly (Johansson & Westling, 1984; Nowak & Hermsdörfer, 2003).

The occurrence of prediction errors when lifting objects of similar visual appearance, but different mechanical properties, is well known since the early descriptions of Johannson, Westling and co-workers (Johansson and Westling, 1984, Johansson and Westling, 1988). In contrast, the ability to use arbitrary visual cues to match them with the mechanical properties of objects to be lifted has only recently been discovered (Chouinard, Leonard, & Paus, 2005; Cole & Rotella, 2002; Nowak, Koupan, & Hermsdörfer, 2007). By now, we know that healthy people can use learned associations between arbitrary colour cues and mechanical object features, such as weight and surface friction, to scale grip force in a predictive manner (Cole & Rotella, 2002; Nowak et al., 2007). It is yet unclear, however, if people can use sensory cues of different modalities, such as visual and auditory stimuli, to learn such associations. In addition, it is unknown if people establish a memory link between an arbitrary sensory cue and the weight of an object to be lifted, regardless of the hand performing the grip-lift task.

Here we investigate the effect of employing sensory cues, such as arbitrary visual (different colours) or auditory cues (different tones), when grasping and lifting an object. Healthy participants were asked to use the precision grip between index finger and thumb to lift different weights in random order with either hand. In the no cue experiment, a non-informative neutral visual stimulus was presented prior to each lift thereby not allowing any judgement about which weight to be lifted. In the cue experiments, either an arbitrary colour cue or an arbitrary auditory cue provided advance information about which of the two weights participants would have to lift in the subsequent trial. Based on earlier data suggesting rapid learning of associations between arbitrary visual cues and mechanical object properties (Chouinard et al., 2005; Cole & Rotella, 2002; Nowak et al., 2007) we hypothesise that healthy people are able to use sensory cues of different modalities to predict the weight of objects to be lifted. From the observations that (i) healthy participants scale grip force precisely to the mechanical object properties within a few lifts, regardless of the performing hand (Johansson and Westling, 1984, Johansson and Westling, 1988), and (ii) the acquired information related to the mechanical object features is easily transferred in between both hands (Gordon et al., 1994), we expect that predictive scaling of grip force based on learned associations is rapidly established, regardless of the hand performing the task.