Transfer of Tactile Learning to Untrained Body Parts: Emerging Cortical Mechanisms

Abstract

Pioneering investigations in the mid-19th century revealed that the perception of tactile cues presented to the surface of the skin improves with training, which is referred to as tactile learning. Surprisingly, tactile learning also occurs for body parts and skin locations that are not physically involved in the training. For example, after training of a finger, tactile learning transfers to adjacent untrained fingers. This suggests that the transfer of tactile learning follows a somatotopic pattern and involves brain regions such as the primary somatosensory cortex (S1), in which the trained and untrained body parts and skin locations are represented close to each other. However, other results showed that transfer occurs between body parts that are not represented close to each other in S1-for example, between the hand and the foot. These and similar findings have led to the suggestion of additional cortical mechanisms to explain the transfer of tactile learning. Here, different mechanisms are reviewed, and the extent to which they can explain the transfer of tactile learning is discussed. What all of these mechanisms have in common is that they assume a representational or functional relationship between the trained and untrained body parts and skin locations. However, none of these mechanisms alone can explain the complex pattern of transfer results, and it is likely that different mechanisms interact to enable transfer, perhaps in concert with higher somatosensory and decision-making areas.

Keywords: perceptual learning; plasticity; somatosensory cortex; somatotopy; specificity; tactile learning; transfer.

Figure 1.

Cortical areas involved in tactile learning and transfer. Area boundaries are derived from cortical parcellations proposed by Van Essen (2005) and Glasser and others (2016). Areas are shown on the inflated left hemisphere of a template brain. Gyri are shown in light gray and sulci in dark gray. Results in monkeys suggest that area 3b should be considered the primary somatosensory cortex (S1 for short; Kaas and others 1979). Eickhoff and others (2010) proposed that somatosensory areas identified in macaque monkeys—including the secondary somatosensory cortex, the ventral somatosensory area, and the parietal ventral area (Disbrow and others 2003; Krubitzer and others 1995)—correspond to subregions of the human parietal operculum, referred to as OP1, OP3, and OP4, respectively. OP2, another subregion of the parietal operculum is part of the vestibular cortex (Eickhoff and others 2010; Frank and Greenlee 2018). Frontal and posterior parietal areas might be involved in transfer if tactile training involves learning to make decisions about tactile stimuli (Pleger and Villringer 2013; Romo and de Lafuente 2013).

Figure 2.

Transfer of tactile learning between trained and untrained fingers of the same hand. (A) Experimental design. The tip of the ring finger of the right hand is trained in a tactile learning task. During the pre- and posttest, performance in the tactile learning task or a related task involving the trained tactile feature is measured by using the tips of the ring finger, adjacent middle finger, and nonadjacent index finger. (B) Theoretical learning and transfer results. Pretest performance is similar among fingers, but posttest performance follows a somatotopic transfer pattern, as evidenced by few errors being made with the trained ring finger and the adjacent untrained middle finger but more errors with the nonadjacent untrained index finger. These results would suggest that tactile learning involved cortical areas with a somatotopic body map in which the middle finger is represented closer to the ring finger than the index finger. For results supporting such a behavioral pattern of transfer of tactile learning, see Harris and others (2001). (C) Cortical representations of the tips of the right ring, middle, and index fingers in a sample participant. Each fingertip was mapped with tactile stimulation during functional MRI: the representation of each fingertip was calculated by contrasting activation during stimulation of this fingertip with activation during stimulation of the other fingertips. The representations of the fingertips are shown on the participant’s inflated left hemisphere by color-coded outlines. Approximate borders among areas 3a, 3b, 1, and 2 (dashed lines) are derived from the cortical parcellation proposed by Glasser and others (2016).

Figure 3.

Tactile learning and transfer between trained and untrained body parts reported by Volkmann (1858; Table XV in his original publication). Volkmann placed a compass circle with two points on a given skin location and reported whether he sensed one or two points. In this experiment, Volkmann used a fixed distance between the two points of the compass circle for each body part (left middle finger, right middle finger, left forearm; tested skin locations are shown by differently colored dots in left panel) and measured how often he sensed two points in a run of 25 trials. Each dot in the right panel shows the result of a different run as a percentage response error that corresponds to the number of times that he sensed only one point in this run. The left middle finger was trained, and transfer of tactile learning to the untrained right middle finger and left forearm was examined. A pretest conducted prior to the beginning of training showed that task performance was similar with trained and untrained body parts. Transfer of learning was examined in the middle and end of training (corresponding to Posttest1 and Posttest2, respectively). The results showed that tactile learning completely transferred from the trained left middle finger to the untrained right middle finger. No transfer of tactile learning was found to the untrained left forearm.

Figure 4.

Mechanisms proposed to explain the transfer of tactile learning between trained and untrained body parts and skin locations. Blue and red blobs correspond to cortical representations of two different body parts labeled A and B. Blue is the trained body part. Red is the untrained body part. (A) Transfer due to adjacent and partially overlapping representations of trained and untrained body parts in S1. (B) Modulation of transfer with the expansion of the representation of the trained body part into the representation of the untrained body part in S1. (C) Transfer due to the latent representation of the untrained body part within the representation of the trained body part and vice versa in S1 (signified by red and blue highlighted rectangles, respectively). (D) Transfer due to functionally coupled representations of trained and untrained body parts in S1. (E) Transfer due to coactivation of the representation of the untrained body part with higher tactile spatial resolution than the trained body part in S1. (F) Transfer due to projections from S1 to higher areas with overlapping or common representations of trained and untrained body parts.

Figure 5.

Distribution of tactile input across representations of stimulated and nonstimulated body parts in the somatosensory cortex (from Frank and others 2022). (A) Tactile stimulation conditions for the palm of the right hand. Tactile movement patterns consisting of four stimuli moving in “v”-shaped (top) and inverted “v”-shaped (bottom) trajectories from left to right were presented to the skin surface through air jets. (B) Tactile stimulation conditions for the sole of the right foot. Otherwise, same as in panel A. (C) Bird’s-eye view of the inflated left hemisphere of the same template brain as in Figure 1. Approximate boundaries among areas 3a, 3b, 1, and 2 (derived from Glasser and others 2016) are indicated by dotted lines. Red and blue colors correspond to the cortical representations of the right hand and foot, respectively, across 16 participants. Regions of greater overlap across participants are shown in more saturated colors. (D) Mean decoding accuracy of functional MRI activation patterns corresponding to “v”-shaped and inverted “v”-shaped tactile movement patterns in the cortical representation of the hand during tactile stimulation of the hand and foot across participants from panel C. Chance level of decoding accuracy corresponds to 50% on the y-axis. (E) Same as in panel D but for the cortical representation of the foot.

Figure 6.

Asymmetrical transfer of tactile learning (from Frank and others 2022). (A) Design of tactile learning with the palm of the right hand and transfer test with the untrained sole of the right foot. (B) Mean learning and transfer results across six participants recruited for the study in panel A. (C) Design of tactile learning with the sole of the right foot and transfer test with the untrained palm of the right hand. (D) Mean learning and transfer results across six new participants recruited for the study in panel C.