Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

Authors: Leeanne M. Carey, PhD Thomas A. Matyas, PhD Stroke Affiliations: From the National Stroke Research Institute, Heidelberg West, Victoria, Australia (LMC); and the Schools of RESEARCH ARTICLE Occupational Therapy (LMC) and Psychological Sciences (TAM), LaTrobe University, Bundoora, Victoria, Australia. Training of Somatosensory Disclosures: Discrimination After Stroke Supported, in part, by a LaTrobe University Faculty of Health Sciences Facilitation of Stimulus Generalization Research Grant, an Australian Research Council small grant, a Research Development Grant from the Department of Health and ABSTRACT Human Services, and a National Health and Medical Research Council Carey LM, Matyas TA: Training of somatosensory discrimination after stroke: Facil- of Australia project grant (191214). itation of stimulus generalization. Am J Phys Med Rehabil 2005;84:000–000.

Presented, in part, at the 11th Objective: Task-specific learning typifies perceptual training but limits rehabilitation of International Congress of the World sensory deficit after stroke. We therefore investigated spontaneous and procedurally Federation of Occupational facilitated transfer of training effects within the somatosensory domain after stroke. Therapists, London, UK, and at the Third Annual Perception for Action Design: Ten single-case, multiple-baseline experiments were conducted with Conference, Melbourne, Australia. stroke participants who had impaired discrimination of touch or limb-position sense. Each experiment comprised three phases: baseline, stimulus-specific Correspondence: training of the primary discrimination stimulus, and either stimulus-specific train- All correspondence and requests for ing of the transfer stimulus or stimulus-generalization training. Both the trained reprints should be addressed to and transfer stimuli were monitored throughout using quantitative, norm-refer- Leeanne M. Carey, PhD, National Stroke Research Institute, Level 2, enced measures. Data were analyzed using individual time-series analysis and Neurosciences Building, Austin meta-analysis of intervention effects across case experiments. Health, Banksia Street, Heidelberg West, Victoria, Australia 3081. Results: Stimulus-specific training was successful for trained texture and proprio- ceptive discriminations, but it failed to show spontaneous transfer to related untrained 0894-9115/05/8405-0001/0 stimuli in the same modality in seven of eight experiments in which this was possible. American Journal of Physical In contrast, intramodality transfer was obtained with stimulus-generalization training in Medicine & Rehabilitation Copyright © 2005 by Lippincott four of five experiments that investigated stimulus-generalization training of texture Williams & Wilkins discrimination. Findings were confirmed by meta-analysis.

DOI: 10.1097/01.PHM.0000159971.12096.7F Conclusions: Our findings demonstrate generalization of training within a somato- sensory modality poststroke, provided that a program designed to enhance transfer is used. This has implications for the design of efficient rehabilitation programs. Key Words: Sensation, Rehabilitation, Cerebrovascular Accident, Transfer of Training

May 2005 Training of Somatosensory Discrimination 1

●●●؍PPL ●●●؍PPF 5؍ISS 84؍May 2005 VID؍Stroke DATE؍Research Articles DOCSUBJ؍ARTICLE DOCTOPIC>

tasks and processing activities are highly similar, it Generalization of intervention effects to un- is unknown whether these conditions are adequate trained tasks is central to sensorimotor neuroreha- for individuals in whom the perceptual system is bilitation. Task-specific training, although effec- damaged. Furthermore, the potential for transfer tive, has the potential to be very costly, and novel of training under different training conditions has tasks will present continuing problems. Discovery not been systematically investigated within the so- of effective training methods also able to achieve matosensory domain after stroke. transfer of gains to novel tasks is thus essential. We have focused our research on rehabilitation Furthermore, investigation of the boundaries of of somatosensory deficits, particularly touch and transfer is increasingly regarded as a key ingredient proprioceptive discrimination. These deficits are to a proper understanding of learning effects in the characteristic of the loss experienced and are re- sensorimotor system.1 In the sensory domain, in- ported in approximately 50% of stroke patients.12 vestigation of stimulus-generalization effects could Such deficits pose substantial difficulties in recep- help clarify what is encoded about the stimuli dur- tion of sensory information and exploration of the ing the training experience. environment and have detrimental effects on spon- “Generalization of training” or “transfer of taneous use of hands, object manipulation, and learning” to an unpracticed task has been investi- precision grip. Moreover, sensory loss has a nega- gated with healthy subjects in various domains, tive effect on personal safety, functional outcome, including perception.2 Empirical investigations in and quality of life, and it influences length of stay the perceptual learning literature suggest that in the hospital (see Carey12 for review), highlight- transfer is possible in some instances.3,4 However, ing the functional importance of training these discrimination training is often highly specific to capacities. the task.2,3,5 A general finding is that similarity This investigation of generalization effects fol- between the trained and transfer tasks is an impor- lowed our initial findings that stroke-induced im- tant determinant of positive transfer.2 pairments of texture discrimination and limb-posi- The dimensions of similarity that seem to mat- tion senses are trainable.8 Improvements were ter include whether the novel transfer task is specific to the stimuli trained2 and confirmed the within the same perceptual dimension as the expected independence of tactile and propriocep- trained task, the similarity of the required re- tive performance.8 Although lack of transfer across sponse, and the similarity in method of processing tactile and proprioceptive discrimination was ex- the information.2–4 Body location, density of recep- pected, those data do not address transfer across tors, and organization of the somatosensory cortex stimuli of the same type, such as textures with may also affect transfer of learning in somatosen- different distinctive features of roughness. Efficient sory discrimination.2–4 The training conditions retraining of somatosensory discrimination abili- under which transfer might be expected also need ties requires an understanding of the similarity to be considered. Principles that may enhance required to achieve successful transfer. transfer of training have been identified in the We therefore conducted a series of ten con- motor-learning literature.1 However, exposure and trolled, multiple-baseline, single-case experi- feedback may be adequate for perceptual transfer, ments13 to investigate transfer across stimuli at least for unimpaired subjects.3–6 within the same sensory-perceptual dimension us- Generalization of training within the same ing either SST or stimulus-generalization training perceptual dimension or sensory modality has re- (SGT). The design permitted training the primary ceived only limited investigation in stroke patients. discrimination stimuli while simultaneously mon- Spontaneous improvement in related visual func- itoring the influence of this treatment on the tions was found in a controlled study that trained a transfer stimuli, and the design delayed introduc- single visual function of patients with cerebral tion of generalization-enhanced training to after blindness.7 In contrast, we found stimulus-specific this control phase. Study 1 investigated spontane- training (SST) effects across tactile and propriocep- ous transfer of training effects within tactile and tive discrimination tasks poststroke.8 Studies that proprioceptive domains after SST. Study 2 investi- have trained multiple modalities simultaneously, gated transfer to novel stimuli within the tactile used objects with multisensory demands,9 or em- dimension using a program modified to facilitate ployed sensorimotor tasks10,11 have reported im- SGT. provement in untrained sensory tasks, suggesting Transfer tasks were selected to require the the potential for generalized somatosensory effects. same sensory-perceptual dimension and body loca- These results are consistent with findings of spon- tion as the trained task, as these features of simi- taneous transfer across multidimensional tasks in larity seem to be important in perceptual learn- healthy subjects.2 Although spontaneous transfer ing.2,3 Tactile stimuli were two types of textured may be achieved by unimpaired subjects when surfaces: plastic grids and fabrics. These stimuli

2 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S require processing within the same tactile domain lidity and normative standards have also been es- and are both explored with the fingertip but have tablished for these tests.14,16 different surface characteristics. The propriocep- tion tasks involved discrimination of imposed wrist Procedure positions in the flexion-extension and ulnar-radial Five single-case experiments were conducted deviation planes of movement. They required the to investigate spontaneous transfer of training to same type of limb-position discrimination and the related stimuli in study 1. Two experiments inves- same body location (i.e., the wrist), but they used tigated transfer within the tactile-discrimination positions within different planes of movement. dimension and three within proprioception. In the Norm-referenced, quantitative, and reliable mea- tactile experiments, texture grids were employed as sures of these tactile and proprioceptive discrimi- the primary stimuli to be trained and the fabrics as nations were available as outcome measures.12,14,15 the transfer stimuli. Imposed wrist positions within the flexion-extension plane were the pri- MATERIALS AND METHODS mary trained proprioception stimuli and positions Study 1: Spontaneous Transfer of within the ulnar-radial deviation plane the transfer Training Effects with SST stimuli. Subjects The typical case experiment had three phases, Five stroke patients with tactile or propriocep- each phase comprising ten sessions in which both tive discrimination impairment, as tested by the the trained and transfer tasks were assessed. As- measures described below, were studied. They were sessment sessions were scheduled 48–72 hrs apart. medically stable, had adequate comprehension of In the first phase, both tasks were monitored only. instructions for assessment, had no peripheral In the second phase, the texture grid or flexion- neuropathy or previous central nervous system extension stimuli were trained over ten treatment dysfunction, and were free of unilateral spatial ne- sessions, interspersed between assessment ses- glect based on clinical observation and standard sions, while baseline monitoring continued on the neuropsychological tests. Patients meeting these transfer response. In the third phase, treatment criteria were selected sequentially, as they pre- was introduced for the transfer stimuli over ten sented, and gave voluntary informed consent. The sessions. This permitted investigation of the sub- project was approved by the human ethics commit- ject’s potential for a SST effect on the transfer task. tees of LaTrobe University and participating hospi- Follow-up was conducted after an interval equiva- tals and conformed to the Helsinki Declaration. lent to the combined time of baseline and inter- vention phases (i.e., 12–14 wks after the end of Materials training). Measures of touch discrimination were the Tactile Discrimination Test (TDT)15 and the Fabric Test Procedure and Scoring Matching Test (FMT).12 The TDT employed finely The TDT was administered after standard in- graded plastic surfaces marked by ridges at set structions were given.15 Subjects tactually explored spatial intervals. Texture grids were presented in each set of comparison surfaces with their pre- sets of three, with two surfaces identical and one ferred finger and indicated the odd texture in a different. The FMT comprised two identical sets of three-alternative, forced-choice design. Five differ- ten cotton-based fabrics, ranked from smooth to ent triplets, spanning Weber ratios of 0.033–1.0, rough by unimpaired subjects. These measures were each presented ten times in random order to have high retest reliability and good discriminative obtain the discrimination limen. The limen was validity and normative standards.15,16 derived from fitting a cumulative normal distribu- For the proprioception domain, the Wrist Po- tion to the probabilities of correct responses. In the sition Sense Test (WPST) quantified the subject’s FMT, subjects attempted to tactually match each of ability to indicate wrist position after an imposed the ten test surfaces with identical comparison movement in the flexion-extension14 or ulnar-ra- surfaces.12 Fabrics were presented behind a cur- dial deviation12 plane of movement. A box-like ap- tain, using a predetermined random order and a paratus included two protractor scales to indicate standard set of instructions. The preferred finger target and response positions and splints for fore- used in the TDT was also used for the fabric test. arm and hand. The hand splint was attached to a Response and target rank orders were correlated lever allowing freedom of movement at the wrist. with Spearman’s rho to quantify fabric discrimina- The box occluded vision of subject’s wrist position tion ability. Rho values were transformed with and of the examiner’s lever manipulations. Each Fisher’s z, and the scale was inverted by subtract- test comprised 20 predetermined wrist positions. ing scores from the allocated maximum value of High retest reliability and good discriminative va- 3.5 z score to make graphical presentation of ther-

May 2005 Training of Somatosensory Discrimination 3 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

apeutic change consistent in direction with TDT tempt to encourage attention to stimulus differ- scores. ences. Quantitative feedback on accuracy of judg- In the WPST for flexion-extension and ulnar- ments was provided by allowing subjects to visually radial deviation, the examiner moved the subject’s check the defined differences (cross-modality cali- hand, via a lever, to 20 different predetermined bration) and by comparing the sensory experience wrist positions. Average absolute error between using the affected and other hand (within-modality actual and response positions was then calculated calibration). Salient sensory features of stimuli as the index of limb position sense for each test. (e.g., roughness characteristics of contour, spatial density) were highlighted by the trainer, and irrel- SST evant variation de-emphasized, in an attempt to The SST program was designed to maximize sensitize subjects to the distinctive features of dif- improvement of the specific sensory discrimina- ference. Qualitative feedback on how subjects ex- tions trained.8,16 Principles of training included: plored the stimuli (e.g., use of lateral movement of repeated presentation of targeted discrimination the fingertip over the surfaces or imagining a ref- tasks (as learning is reported to be maximal for the erence point of where the hand was pointing in the specific task trained); progression from easy to proprioception task) was also given. After initial more difficult discriminations; attentive explora- exposure to stimuli of a particular stimulus differ- tion of stimuli with vision occluded; use of antici- ence level, a series of anticipation trials were con- pation trials; feedback on salient sensory features ducted in which subjects were instructed that the of the stimuli, accuracy of judgments, and method same stimuli set previously explored would be pre- of exploration; comparison of the sensation with sented. Thus, choices were limited and subjects the other hand; use of vision to facilitate inter- could make judgments with knowledge of what to modal calibration of sensory information; sum- expect to feel. Subjects responded whether the mary feedback; and intensive training.8,16 These stimuli were the same or different and identified principles were applied to training each stimulus, which of the limited stimuli was felt. Feedback on as described below. accuracy, sensation, and method of exploration was In the SST program, training tasks were the provided, as above. When correct discriminations same as the assessment tasks, but typically used a occurred in at least three of four consecutive oc- restricted range and number of stimuli. Five sets of casions, the next-finer stimulus difference was in- grids, with differences in surfaces ranging from troduced. Summary feedback on the individual’s 3.3% spatial increase to 100% spatial increase, accuracy of judgments and method of exploration were used for training texture grids, as previously was also provided at the end of each training ses- described.8 Subjects explored each set of compari- sion. Training was intensive, with sessions lasting son grid surfaces with their preferred finger and 40–60 mins, three times a week. indicated the odd texture. Training of fabric dis- crimination used the fabrics of the FMT. Subjects Data Analysis were required to match comparison surfaces to a First, each single-case experiment was ana- restricted range of target surfaces and, in some lyzed separately for stimulus-specific and general- trials, to identify which fabrics they had touched. ized training effects. Second, these results were Training employing a subset of fabrics with four combined in a meta-analysis to obtain an overall levels of difference, from large to small differences, conclusion. provided a framework to assist interpretation of the Time-series data were analyzed using visual relative smoothness and roughness of the set of (graphical) and statistical analyses.8,13 For graphi- surfaces. To train for flexion-extension position cal analyses, standard single-case charts were eval- sense,8 the subject’s wrist was moved to a position uated visually by the authors and a panel of three in the flexion-extension range. Subjects then indi- trained, independent analysts who were naïve to cated perceived position of the wrist from a set of the purpose of the study. The analysts’ type I error defined angles. Seven positions were used to (false alarms) and type II error (missed treatment achieve four levels of stimulus difference (i.e., 90-, effect) rates were calibrated using computer-gen- 60-, 30-, and 15-degree differences). In training for erated charts in which null or non-null effects were ulnar-radial position sense, six positions were used presented unidentified and in random order.8,16 to achieve four levels of stimulus difference (i.e., Analysts were required to make separate judgments 40-, 25-, 15-, and 7-degree differences). of systematic change between phases according to Subjects were initially presented with the larg- defined criteria. est stimulus differences for which Ͻ100% accuracy Statistical interrupted time-series analyses had been shown in testing during the baseline (ITSA)13 were also conducted for each single-case phase. Stimulus sets were explored with vision experiment. The initial step comprised model iden- occluded, and subjects gave a response in an at- tification, in which three curvilinear models of

4 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S learning likely to adequately describe the interven- Procedure tion effect were compared for goodness of fit using This study comprised five multiple-baseline, the curvilinear regression routines of SPSS. The single-case experiments. Performance on the GMT models Y ϭ a ϩ blnX, Y ϭ aXb and Y ϭ aebX were 8 and FMT was monitored throughout the time se- compared, consistent with earlier investigations. ries. The first phase comprised baseline only. In the Residuals from the models were examined by au- second phase, SST (described in study 1) was in- tocorrelation and partial autocorrelation methods troduced for texture grids while the transfer re- to detect remaining serial dependence in the data. sponse was monitored. In the third phase, SGT was The best-fitting model with residuals free of posi- introduced while monitoring of the untrained tive serial dependence was selected to describe the transfer response continued. Follow-up was con- data and perform ITSA. After investigation and ducted 12–14 wks after the end of training. deletion of outliers, ITSA were tested for trend and level effects between phases to evaluate interven- Test Procedure and Scoring tion effects, as previously described.8 A systematic difference between phases was judged to be present The stimulus-matching procedure was used if either a statistically significant trend or level for both tests. For the GMT, the subject was re- effect was present. Determination of a spontaneous quired to tactually match each of eight test sur- transfer effect involved (a) identification of an in- faces (selected from the set of 16) against the tervention effect on the primary trained response identical set, with vision occluded. For the FMT, together with (b) identification of a simultaneous, subjects were required to match each of the ten systematic change in a therapeutic direction in the test surfaces with the ten comparison surfaces. transfer response. Testing took approximately 10–15 mins for each Finally, meta-analyses were conducted to com- test. Stimulus-matching ability was quantified by plement individual subject ITSAs and assist a gen- correlating the response and test values with Pear- eralized conclusion across the case experiments. son’s r (GMT) and Spearman’s rho (FMT). Corre- The type I error probabilities of each ITSA were lation coefficients were then normalized with Fish- converted to z scores and then averaged to obtain a er’s z transform. A higher score reflects better pooled estimate.17 Separate meta-analyses exam- matching. ined stimulus-specific intervention effects, sponta- neous generalization effects, and the effect of the SGT intervention designed to facilitate generalized im- The SGT program was designed to facilitate provement in texture discrimination. transfer of training effects to untrained, novel stimuli. To achieve this, the additional principles of Study 2: Transfer of Training Across variation in stimulus and practice conditions, in- Textured Surfaces Using a Program termittent feedback, and tuition of training princi- Designed to Facilitate Generalization to ples1,2,19 were included, as these have been associ- Novel Stimuli ated with enhanced transfer and retention. A Subjects variety of training surfaces (e.g., paper, glass, leather, and rubber, but not fabrics) were employed Five further stroke patients, who met selection in the SGT program. The surfaces comprised a criteria previously described and had tactile dis- range of distinctive features of roughness, includ- crimination impairment, were investigated. Pa- ing contour, surface pattern, and grit. Each differ- tients meeting these criteria were selected sequen- ent type of surface was graded from smooth to tially, as they presented, and gave voluntary rough across five stimuli and included small to informed consent. large differences. First, large differences were in- troduced across a subset of surface types, followed Materials by medium and fine differences. Thus, progressive The FMT was again employed, as previously grading of difficulty was across stimuli and within described. The Grid Matching Test (GMT)12 tested stimuli, and subjects had the opportunity of mak- texture-grid matching ability. The test used the ing similar discriminations across novel surface same finely graded plastic grids as for the TDT but types. Feedback on salient sensory features, accu- employed a stimulus-matching procedure, as in racy, and exploration method (as for SST) was the FMT. Sixteen surfaces, with spatial intervals given intermittently rather than at every trial, and ranging from 1500 to 3000 ␮m, were placed on test subjects were encouraged to check the accuracy of and comparison texture wheels and presented for their own performance. Feedback was also given on tactual exploration through openings in the test the transfer task of identifying new distinctive fea- apparatus. Reliability, discriminative validity, and tures of roughness in novel stimuli. Specific tu- normative data have been obtained for this test.18 ition on principles underlying training, such as use

May 2005 Training of Somatosensory Discrimination 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

of anticipation trials and feedback, and how these hot/cold discrimination using the Roylan hot and apply across tasks that the client may encounter in cold discrimination kit (A629-1) (age-matched, other environments were also included. Repeated healthy controls detect at least 9 out of 10 correct16). presentation of stimuli, attentive exploration with vision occluded, anticipation trials, summary feed- Model Identification back, and intensive training were also incorpo- The logarithmic model Y ϭ a ϩ blnX ac- rated, as for SST. counted for most variance in the intervention data for eight of the ten time series, without high pos- Data Analysis itive autocorrelation in the residuals of any series. Case charts were analyzed visually and statis- Statistically significant negative autocorrelation tically for trend and level effects. Individual time (lag 2) was identified only in two baseline phases. series were graphed for visual analysis, and statis- The analysis of residuals thus confirmed selection tical analysis was conducted, after model identifi- of a logarithmic model for ITSA. cation, as previously described. Intervention Effects in the Trained RESULTS Response Study 1 Intervention effects were clearly apparent in T1 Background data of subjects is detailed in Ta- three of the five first-trained primary responses ble 1. All subjects showed marked impairment on after SST (Fig. 1, S01 to S03). These improvements F1 the TDT (i.e., unable to consistently discriminate were detected above any changes between the first the largest texture difference of 100% spatial in- and second phase of baseline of the second discrim- crease [criterion of normality, Ͻ37.7% spatial ination response. Improvements were marked and increase15]). Impaired performance was also immediate, with changes in performance noted evident across multiple modalities, using the fol- after the first training session. Performance scores lowing tests and normative standards: wrist pro- achieved by the end of training were within the prioception using the WPST for flexion-extension normal range15 and similar to those of the other (criterion of normality Ͻ9.5 degrees14,16); pressure hand in all three cases. The panel of judges unan- in (0.00 ف discrimination using Semmes-Weinstein monofila- imously detected a systematic change (P ments (normal threshold ϭ 0.02–0.13 filament20); all three series and did not identify a significant

TABLE 1 Background information of subjects in study 1

Subject

S01 S02 S03 S04 S05

Age, yrs 44 60 59 51 50 Sex Male Male Male Female Male Hand dominancea Left (consistent) Right (consistent) Right (consistent) Right (consistent) Right (consistent) Affected side Left Left Right Right Left Time since stroke 10.5 wks 6.5 wks 5.5 wks 13.5 wks 8 wks onset Site of lesion (based Right frontoparietal Right temporal lobe, corona Left frontoparietal and Left lentiform- Right basal ganglia, on CT scan) infarct radiata, and lentiform corona radiata infarct internal capsular internal capsule nucleus infarct hemorrhage hemorrhage Motor function Limited voluntary movement Isolated finger control No voluntary movement No functional No functional movement (affected UL) of wrist and fingers movement Sensory function TDT ϭ 100 TDT ϭ 100 TDT ϭ 100 TDT ϭ 100 TDT ϭ 100 (affected UL) PSI WPST ϭ 23.9 degrees PSI WPST ϭ 11.3 degrees PSI WPST ϭ 32.4 degrees PSI WPST ϭ 32.7 PSI WPST ϭ 25.2 degrees Pressure ϭ 6.65 Pressure ϭ 4.31 Pressure ϭ 4.31 degrees Pressure ϭ 6.65 H/C ϭ 5/10 H/C ϭ 7/10 H/C ϭ 6/10 Pressure ϭ N/A H/C ϭ 0/10 H/C ϭ N/A Cognitive function Slight decrease in complex Impaired language; integrity Mildly reduced attention NAD; alert; NAD; attention, memory, planning skills, otherwise NAD of problem solving and processes; slow concentration, and learning intact memory information processing memory, and learning intact CT scan, computed tomographic scan; UL, upper limb; TDT, Tactile Discrimination Test15; PSI, percentage of spatial increase; WPST, Wrist Position Sense Test in flexion-extension plane14; Pressure, log10 force of Semmes-Weinstein monofilament (in milligrams) detected at preferred finger17; H/C, hot/cold discrimination using Roylan hot and cold discrimination kit (A629-1), number of correct judgments out of ten stimuli presented to distal pad of preferred finger are reported; N/A, not available; NAD, no abnormality detected. aHand dominance determined from Annett21 questionnaire of hand dominance.

6 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

FIGURE 1 Case charts of spontaneous transfer effects within tactile and proprioceptive domains. In phase 1 (sessions 1–10), performance was monitored under baseline conditions for both trained and transfer stimuli. In phase 2 (sessions 11–20), stimulus-specific training was introduced to the primary trained response (texture grids or flexion-extension wrist positions). Performance of the transfer response (fabrics or ulnar-radial wrist positions) was simultaneously monitored under extended baseline conditions to permit observation of spontaneous transfer effects. In phase 3 (sessions 21–30), stimulus-specific training was introduced to the transfer response and continued for the primary trained response but at a lower intensity. Texture grid limen (PSI), score from the Tactile Discrim- ination Test in units of percentage of spatial increase.15 Fabric matching z-score, score from Fabric Matching Test in Fisher’s z units.12,16 Fisher’s z scale was inverted so that graphical representation of the Fabric score was consistent with the Tactile Discrimination Test score (i.e., higher values represent poorer performance). Flex-ext position error (deg), flexion-extension wrist position sense average error, in degrees, based on the Wrist Position Sense Test.14 Uln-rad position error (deg), ulnar-radial deviation wrist position sense average error, in degrees, based on the Ulnar-Radial Wrist Position Sense Test.12,16 S01–S05, subjects 1–5.

parallel change in the transfer response, suggest- Lack of a concomitant and systematic change in ing a specific training effect. Statistical analysis the transfer responses was confirmed by the panel confirmed both trend and level effects in the three of judges. Although a statistically significant T2 series (Table 2). Improvements achieved by the end change in trend was obtained for subject 3 (Table 3, T3 of the training were well maintained. In the re- S03) the change was not in the therapeutic direc- maining two series, no intervention effect could be tion. Subject 2 demonstrated variable performance claimed (Fig. 1, S04 and S05). Both series showed on the transfer task (FMT) during sessions 1–10 of an improvement during baseline, limiting the baseline and a reduced variability and possible im- scope for training effects. provement during sessions 11–20 of baseline. It was difficult to distinguish whether these observa- Spontaneous Transfer of Training tions represented continuation of the pattern es- Spontaneous transfer of training to the second tablished in the first baseline phase or a change in response was not evident in two of the three cases the second baseline phase; therefore, judgment was in which an intervention effect was present in the reserved. The panel of visual analysts failed to in- primary trained response (Fig. 1, S01 and S03). dicate a systematic change. However, a statistically

May 2005 Training of Somatosensory Discrimination 7 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

TABLE 2 Stimulus-specific training effects: Summary of trend and level effects from interrupted time series analyses Trend Effect Level Effect

Subject Time Series t(df) P, 1-Tail t(df) P, 1-Tail

Study 1 S01 Texture grids t(26) ϭ 5.67 <0.0001 t(26) ϭ 2.63 <0.008 Fabricsa t(15) ϭ 3.10 <0.004 t(15) ϭ 1.43 Ͼ0.05 S02 Texture gridsa t(25) ϭ 8.83 <0.0001 t(25) ϭ 8.85 <0.0001 Fabrics t(16) ϭ 2.67 <0.009 t(16) ϭ 0.97 Ͼ0.05 S03 Flex-ext t(26) ϭ 2.51 <0.01 t(26) ϭ 5.79 <0.0001 Ulnar-radial t(16) ϭ 1.06 Ͼ0.05 t(16) ϭ 4.66 <0.0002 S04 Flex-ext t(26) ϭ 3.22 Ͻ0.002 t(26) ϭ 0.90 Ͼ0.05 Ulnar-radial t(16) ϭ 1.85 <0.04 t(16) ϭ 4.82 <0.0001 S05 Flex-ext t(26) ϭ 4.06 Ͻ0.0002 t(26) ϭ 2.66 Ͻ0.007 Ulnar-radial t(16) ϭ 1.52 Ͼ0.05 t(16) ϭ 3.21 <0.003 Study 2 S06 Texture grids t(16) ϭ 4.57 <0.0002 t(16) ϭ 5.17 <0.0001 S07 Texture grids t(16) ϭ 2.92 <0.005 t(16) ϭ 6.98 <0.0001 S08 Texture grids t(16) ϭ 0.09 Ͼ0.05 t(16) ϭ 3.91 <0.0006 S09 Texture grids t(21) ϭ 3.34 <0.002 t(21) ϭ 0.75 Ͼ0.05 S10 Texture grids t(18) ϭ 3.84 <0.0006 t(18) ϭ 0.45 Ͼ0.05 Flex-ext, wrist position stimuli in the flexion-extension plane of movement; ulnar-radial, wrist position stimuli in the ulnar-radial deviation plane of movement. aOutliers deleted. Bold P values indicate a significant effect in a therapeutic direction; italic P values (S04 and S05) indicate a significant difference between phases, but an intervention effect was not claimed due to pre-existing improvement in baseline. Comparisons are between baseline (second phase of baseline for transfer stimuli) and stimulus-specific training phases.

TABLE 3 Spontaneous transfer effect to the transfer discrimination response Trend Effect Level Effect

Subject Time Series t(df) P, 1-Tail t(df) P, 1-Tail

Study 1 Tactile discrimination S01 Fabrics t(16) ϭ 0.60 Ͼ0.05 t(16) ϭ 0.51 Ͼ0.05 S02 Fabrics t(16) ϭ 0.38 Ͼ0.05 t(16) ϭ 2.44 <0.01 Proprioception discrimination S03 Ulnar-radial t(16) ϭ 2.29 Ͻ0.02 t(16) ϭ 0.45 Ͼ0.05 S04 Ulnar-radial t(16) ϭ 1.95 Ͻ0.04 t(16) ϭ 0.93 Ͼ0.05 S05 Ulnar-radial t(16) ϭ 0.26 Ͼ0.05 t(16) ϭ 0.59 Ͼ0.05 Study 2 Tactile discrimination S06 Fabrics t(16) ϭ 1.32 Ͼ0.05 t(16) ϭ 0.15 Ͼ0.05 S07 Fabrics t(16) ϭ 1.41 Ͼ0.05 t(16) ϭ 1.01 Ͼ0.05 S08 Fabrics t(16) ϭ 1.20 Ͼ0.05 t(16) ϭ 0.23 Ͼ0.05 S09 Fabrics t(21) ϭ 2.32 <0.02 t(21) ϭ 0.23 Ͼ0.05 S10 Fabrics t(18) ϭ 0.37 Ͼ0.05 t(18) ϭ 0.08 Ͼ0.05 Ulnar-radial, wrist position stimuli in the ulnar-radial deviation plane of movement. Bold P values indicate a significant effect in a therapeutic direction; italic P values indicate a significant difference between phases but not in a therapeutic direction. Comparisons are between the first and second phase of baseline for the transfer stimulus.

significant change in level, from 2.6 to 2.1, on the identified; therefore, generalization of a training fabric-matching score was observed (Table 3). The effect was not testable. identified change was in a therapeutic direction but of small magnitude and did not seem to follow the Effect of SST on the Transfer Stimuli characteristic logarithmic function. In the remain- Although the fabric and ulnar-radial position ing two cases (subjects 4 and 5), an intervention stimuli did not demonstrate evidence of spontane- effect in the first-trained primary response was not ous generalization effects, they did respond to SST

8 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

in all cases (Fig. 1). Improvements were clearly Effects of SST defined and marked relative to the second phase of The case charts (Fig. 2) suggested a systematic F2 baseline. All subjects performed within the normal change between baseline and SST phases for the range on the FMT and WPST for ulnar-radial devi- 16 GMT beyond any parallel change in the transfer ation by the end of training. In the ulnar-radial responses of each of the subjects. Improvements series, the lowest error scores were achieved. Inde- were marked, often achieving scores within the ف pendent visual analysis (P 0.00) and statistically normal performance range18 and comparable with significant trend and level effects (Table 2) con- the other hand. These effects of SST were con- firmed the presence of intervention effects relative firmed for each subject by the ITSAs (Table 2, S06 to the control baseline. Improvements achieved by to S10). the end of training were well maintained at fol- low-up in all cases. Spontaneous Transfer to Untrained Fabric Study 2 Stimuli After SST Background details of subjects are presented in Spontaneous transfer to the untrained FMT T4 Table 4. Subjects showed severe deficit on the GMT response was not evident in the second baseline (-0.12 to 0.29 z score compared with 0.62 z score, phase when SST of texture grids was introduced the criterion of normality18) and moderate to se- (Fig. 2). This was clear for subjects 6–8. Subjects 9 vere impairment on the FMT (criterion of normal- and 10 showed more variability. Lack of a transfer ity, Ͼ1.61 z score18). Most patients had the maxi- effect was confirmed by the results of the ITSA in mum impairment score on the TDT. Performance all but one case (Table 3). A significant trend effect on the WPST varied. was identified in subject 9. However, the change did not follow the characteristic pattern achieved Model Identification by training, and performance was variable; thus, a The logarithmic model previously identified in spontaneous transfer effect was not claimed. several of our studies8,16 and in study 1 again successfully accounted for baseline and interven- Transfer to Novel Textured Stimuli After tion data, leaving no statistically significant or sus- SGT piciously high positive autocorrelation in the re- Discrimination of fabric surfaces, which had siduals of any time series. not been trained in any of the three phases, im-

TABLE 4 Background information of subjects in study 2

Subject

S06 S07 S08 S09 S10

Age, yrs 63 88 70 65 47 Sex Female Female Male Male Male Hand dominancea Right (consistent) Right (consistent) Left (inconsistent) Right (inconsistent) Right (consistent) Affected side Right Left Right Left Right Time since stroke 5 wks 4.5 wks 13 wks 52 wks 18 wks onset Site of lesion (CT Left striatocapsular Right posterior parietal Left posterior limb Infarct in right Extensive left scan) infarct and basal ganglia internal capsular occipital and parietal temporoparietal and infarct hematoma (superior) lobes corona radiata infarct Motor function No active movement Limited active No functional use Independent finger Independent finger (affected UL) movement; (flicker only) movement; movement Preexisting tremor clumsy Sensory function GMT ϭ 0.15 GMT ϭ 0.05 GMT ϭϪ0.12 GMT ϭ 0.18 GMT ϭ 0.29 (affected UL) FMT ϭ 0.56 FMT ϭ 0.12 FMT ϭϪ0.16 FMT ϭ 0.99 FMT ϭ 0.89 TDT ϭ N/A TDT ϭ 100 TDT ϭ 100 TDT ϭ 100 TDT ϭ 100 PSI WPST ϭ N/A PSI WPST ϭ N/A PSI WPST ϭ 34.4 PSI WPST ϭ 11.6 PSI WPST ϭ 13.2 degrees degrees degrees Cognitive NAD NAD Expressive dysphasia NAD Dysphasia; function mild visuospatial problems CT scan, computed tomographic scan; UL, upper limb; GMT, Grid Matching Test (Fisher’s z scores)12; FMT, Fabric Matching Test (Fisher’s z scores)12; TDT, Tactile Discrimination Test15; PSI, percentage of spatial increase; WPST, Wrist Position Sense Test in flexion-extension plane14; N/A, not available; NAD, no abnormality detected. aHand dominance determined from Annett21 questionnaire of hand dominance.

May 2005 Training of Somatosensory Discrimination 9 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

FIGURE 2 Case charts of stimulus-generalization training effects in the tactile domain. In phase 1 (typically sessions 1–10), performance was monitored under baseline conditions for both trained and transfer stimuli. In phase 2 (typically sessions 11–20), stimulus-specific training (SST) was introduced to the texture grids and spontaneous transfer effects were monitored in the transfer-fabric stimulus. In phase 3 (typically sessions 21–30), stimulus-generalization training (SGT) was introduced, and monitoring continued for the fabric response to permit observation of transfer of training effects. Grid-matching and fabric-matching test scores are in Fisher’s z units.12,16 In contrast to Figure 1, higher values represent better performance. Case chart S06:trained/untrained grids provides an example of the case charts in which the grid-matching score was separately calculated for grids that were trained and those that were not trained. Stimulus-specific training effects are evident in the trained grid score, and spontaneous transfer of training effects are evident in the untrained grid score. S06–S10, subjects 6–10.

proved on introduction of the SGT program in four achieving scores within the normal performance of the five experiments (subjects 6–9), as shown in range.18 They followed a pattern similar to the SST Figure 2. Improvements were often marked, effects, although the magnitude of early change

10 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

was often smaller. Subject 10 also showed an im- 0.126, P Ͼ 0.55), despite the increase in sensitivity provement trend in phase 3; however, this was not afforded by the pooling of seven sets of experimen- significantly different from the preexisting trend tal results. A meta-analysis that directly contrasted T5 evident in phase 2 (Table 5). Statistical analyses are the effect of specific grid training on discrimina- summarized in Table 5. tion of plastic grids vs. discrimination of fabrics showed a clearly superior effect in the discrimina- Spontaneous Transfer from Trained to tion of the plastic grids according to both level (Z Untrained Grids After SST ϭ 5.24, P Ͻ 0.001) and trend (Z ϭ 6.07, P Ͻ 0.001) To investigate whether transfer might occur variable results. Importantly, when the SGT pro- across stimuli with the same texture characteristic, gram was introduced in the second group of five we trained for only half of the grid surfaces, while experiments, both level (Z ϭ 4.10, P Ͻ 0.001) and monitoring performance on all (Fig. 2, S06: trend (Z ϭ 5.57, P Ͻ 0.001) variables responded trained/untrained grids). Spontaneous transfer with significant changes in the untrained transfer from trained texture grids to untrained texture stimuli. grids did occur with the SST program in four of the T6 five case experiments (Table 6). Subject 10 was the DISCUSSION only subject who did not show either spontaneous Several conclusions seem reasonable. First, transfer to stimuli with the same distinctive feature rapid, task-specific training effects can be obtained (texture grids) or facilitate transfer to novel stimuli with the SST program. Second, improvements ob- (fabrics). tained in a particular discrimination task do not generalize spontaneously to related stimuli within Meta-analyses the same sensory-perceptual dimension in stroke A meta-analysis investigating the overall effect patients after SST. For example, training on grid of SST with the plastic texture grids was performed surfaces did not generalize to fabrics. However, on the seven subjects providing appropriate data spontaneous transfer of training does seem to oc- (subjects 1, 2, and 6–10). This meta-analysis very cur across stimuli that share the same distinctive clearly confirmed the positive effect of training on feature (e.g., trained and untrained grids) after the discrimination of texture grids in both level (Z SST. Third, performance on related untrained ϭ 7.61, P Ͻ 0.001) and trend (Z ϭ 8.46, P Ͻ 0.001) stimuli within the same sensory-perceptual dimen- variables. Discriminations within the same diffi- sion can be improved to a clinically significant culty range, but using plastic grids with spatial degree by using a program deliberately oriented to intervals that were not used during training trials, enhance generalization (the SGT program). Trans- were available from five experiments (subjects fer of training from practice with paper, glass, and 6–10). This meta-analysis also showed significantly rubber to discrimination of fabric surfaces illus- improved performance in both the level (Z ϭ 3.43, trates this point. P Ͻ 0.001) and trend (Z ϭ 4.61, P Ͻ 0.001) variables. These strong positive results in discrim- Conditions of Training and Influence on inations of trained and untrained texture grids Transfer contrast markedly with the results obtained for Successful transfer across stimuli seems to spontaneous transfer–related changes in the dis- have been influenced by the training principles crimination of fabric textures after SST of plastic employed. The SST program employed principles grids. Meta-analysis of these time series failed to designed to maximize learning to discriminate a show a statistically significant effect in either level particular stimulus type. Positive training and (Z ϭ -0.206, P Ͼ 0.41) or trend changes (Z ϭ transfer effects within the specific stimulus type

TABLE 5 Stimulus-generalization training effect: Transfer to novel textured stimuli Trend Effect Level Effect

Subject Time Series t(df) P, 1-Tail t(df) P, 1-Tail

S06 Fabrics t(16) ϭ 5.37 <0.0001 t(16) ϭ 3.76 <0.001 S07 Fabrics t(16) ϭ 3.41 <0.002 t(16) ϭ 1.59 Ͼ0.05 S08 Fabrics t(16) ϭ 4.13 <0.0004 t(16) ϭ 2.22 <0.02 S09 Fabrics t(21) ϭ 1.77 Ͼ0.05 t(21) ϭ 2.10 <0.02 S10 Fabrics t(23) ϭ 1.13 Ͼ0.05 t(23) ϭ 1.31 Ͼ0.05 Bold P values indicate a significant effect in a therapeutic direction. Comparisons are between the second phase of baseline and the stimulus-generalization training phase for the transfer stimulus.

May 2005 Training of Somatosensory Discrimination 11 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

TABLE 6 Stimulus-specific training effect on trained grids and spontaneous transfer to untrained texture grids Trend Effect Level Effect

Subject Texture Grids t(df) P, 1-Tail t(df) P, 1-Tail

S06 Trained t(16) ϭ 5.49 <0.0001 t(16) ϭ 6.66 <0.0001 Untrained t(16) ϭ 3.02 <0.004 t(16) ϭ 1.34 Ͼ0.05 S07 Trained t(16) ϭ 1.49 Ͼ0.05 t(16) ϭ 4.70 <0.0001 Untrained t(16) ϭ 0.62 Ͼ0.05 t(16) ϭ 3.21 <0.003 S08 Trained t(16) ϭ 0.67 Ͼ0.05 t(16) ϭ 2.72 <0.0002 Untrained t(16) ϭ 0.41 Ͼ0.05 t(16) ϭ 2.31 <0.02 S09 Trained t(21) ϭ 4.77 <0.0001 t(21) ϭ 1.08 Ͼ0.05 Untrained t(21) ϭ 5.04 <0.0001 t(21) ϭ 0.12 Ͼ0.05 S10 Trained t(18) ϭ 5.30 <0.0001 t(18) ϭ 0.06 Ͼ0.05 Untrained t(18) ϭ 0.40 Ͼ0.05 t(18) ϭ 0.66 Ͼ0.05 Bold P values indicate a significant effect in a therapeutic direction. Comparisons are between baseline and stimulus-specific training phases for the trained grids and between the first and second phase of baseline for the untrained grids.

trained were achieved. However, spontaneous The high degree of specificity observed is also transfer to related tasks within the same sensory- consistent with evidence of highly specific deficits perceptual dimension did not occur. To facilitate resulting from cerebral damage.12,21,22 Moreover, generalized improvements, a program that added studies of behaviorally induced neural plasticity variation in training stimuli, intermittent feed- suggest that changes in cortical maps are specific back, and tuition of training principles was re- to the trained task and specific body location in quired. Repeated exposure under conditions of at- monkeys23 and humans.24 Organization of the so- tentive exploration with vision occluded was matosensory system is highly specific, as evidenced usually insufficient for therapeutic change, as in- by body location and modality-specific columnar dicated by stable baseline performance in 75% of organization and submodality-specific neurons in the baseline time series investigated. Thus, condi- the primary somatosensory cortex and neuro-ax- tions of training seem important to outcome, with is.25 These observations indicate a high degree of implications for the design of efficient rehabilita- specificity of information processing within modal- tion programs. The principles of training derived ities and body locations. However, the sensory sys- from studies of learning with normal subjects seem tem is also characterized by convergence of so- applicable to stroke patients, confirming the value matosensory information26 and presence of of learning theory for models of neurologic distributed sensory networks.25 Thus, distinctive rehabilitation. sensory features of a multidimensional texture might be integrated in a unified perception at a Specificity of Learning and Information level of convergence in the system.26 This level of Processing Within the Sensory System processing would be consistent with the learning Perceptual learning studies with unimpaired and transfer across textures with multiple features subjects suggest that discrimination training is of- of roughness after SGT. ten highly specific to the task and receptor location and to the method of processing information.2,3 We What Is Learned also found highly specific training effects in tasks A common view supported by behavioral2 employing the same sensory dimension (tactile or and neurophysiologic27 evidence proposes that proprioceptive) and body location (fingertip or distinctive features of difference are learned and wrist) when SST was undertaken. Spontaneous form the basis of transfer of training. In the TDT transfer was, however, found across stimuli of the and GMT, the distinctive feature is likely to be same type that employed the same receptor loca- the spatial interval of the grids,27 whereas the tion and method of processing the information physical characteristics of fabrics are likely to be (e.g., from trained to untrained grids and across based on nonspatial neural coding mecha- wrist positions within the same plane of movement nisms.27 This inference is supported by observa- [positions trained were different to those as- tion of stimulus-specific transfer within, but not sessed]). These findings suggest spontaneous across, these stimuli. The distinctive feature of transfer of training to novel stimuli of the same the limb-position task is presumed to be the type and of comparable difficulty when SST is change in location of the limb28 or relative wrist employed. angle. Our findings suggest that the distinctive

12 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S feature also needs to be defined in relation to the provide some guidelines on what features of the defined plane of movement or to receptors in- task need to be similar before learning transfers. volved in the discrimination. This conclusion is Subjects with an unimpaired sensory system consistent with observations indicating selective can improve perceptual discriminations with prac- activation of muscle receptors,29 joint receptors, tice alone and demonstrate spontaneous perceptual and skin areas in association with specific posi- transfer.3 In contrast, stroke patients in the present tions.28 Thus, it seems necessary that the distinc- studies repeatedly practiced with the assessment tive features of stimulus difference are highly stimuli during baseline, yet in the majority of similar and specific to the receptor population cases, they did not demonstrate improvement on activated to obtain spontaneous generaliza- these stimuli. This finding highlights the need tion. for appropriate conditions in rehabilitation training. Using the SGT program, transfer was obtained Repeated exposure alone, which approximates across tactile stimuli with potentially different dis- exposure to sensory stimuli in daily activities, tinctive features (i.e., from training on rubber, was not adequate to effect a clinically significant glass, leather, and sandpaper to transfer on fab- improvement. rics). Training of multidimensional stimuli with a After damage to somatosensory cortical areas wide range of distinctive features in the SGT pro- and their connections, it is likely that stroke pa- gram is likely to increase the probability of expo- tients have an inadequate foundation from which sure to distinctive features relevant to the transfer to perceive and discriminate novel stimuli, al- stimulus, with consequent transfer of learning. In- though they may have learned the principles of vestigations with unimpaired subjects have sug- how to process similar types of information better. gested that transfer is facilitated when stimuli are Perceptual learning at any given time is con- 2 more complex and potentially share some distinc- strained by the existing structure. Evidence that tive features.2 Physical characteristics of difference learning involves modification of representations 23,24 trained in our study included surface type, surface in the brain suggests that perceptual training contour, grit, friction/abrasion, stimulus patterns is not only a means of training the process of with different spatial frequencies, surface irregu- perceptual learning. Individuals must have access larities, and various combinations of these. This set to the distinctive characteristic of the stimulus to of features seems to represent a broader dimen- extract the distinctive feature. External quantita- sion2 of roughness and covers aspects of roughness tive feedback on what is being perceived may be that are both spatial and nonspatial. Thus, success necessary to make sense of or to “calibrate” per- of the SGT may be influenced by the fact that the ceptions in a changed (e.g., after brain damage) 28 varied stimuli used included nonspatial discrimi- system. nations that may be more directly related to the Study Limitations transfer stimulus. Further, it may be the dimen- sion of roughness, rather than a specific unitary Investigation of training effects was limited to feature of it, that is learned and transferred in this ten subjects (20 time series). Although this is a training. Other studies obtaining generalized train- small number for group-based studies, each inten- ing effects in stroke patients9–11 have included sive single case is a controlled experiment involv- variation in training stimuli, supporting this ing within-subject control and replication of train- interpretation. ing effects. This approach is well suited to the heterogeneous nature of stroke and allows for vari- able findings across individuals that might be ob- Generalization of Training Within a scured with group analysis. In addition, systematic Disordered Somatosensory System replication of training effects across ten subjects The ability of stroke patients to generalize with different background characteristics and learning to novel tasks is relatively unknown, par- across tactile and proprioceptive modalities pro- ticularly within the somatosensory domain. Stroke vides generalizability of findings not usually af- patients with somatosensory impairment may not forded to single case studies. Finally, meta-analysis have the capacity to process the intrinsic somato- of individual experiments improves power, permits sensory input from the untrained transfer task or an overall conclusion, and provides a quantitative may have a more global difficulty in generalizing evaluation of the generality of treatment effects.13 learning effects. Our findings confirm both task- Thus, our findings provide a strong foundation for specific and generalized transfer effects. The ob- larger controlled studies. served effects suggest that stroke patients can The sequence of baseline, SST, and SGT was transfer learning effects, but the boundaries of constant, and therefore, it may be argued that the transfer do have limits. More importantly, they exposure and SST phases led to an overlearning suggest a high degree of specificity of learning and effect with an impact on transfer and effectiveness

May 2005 Training of Somatosensory Discrimination 13 Art: 162531 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

of the SGT program. However, this explanation is sensory function after stroke30 may provide new unlikely given that observed SST effects worked insights into cerebral networks involved in percep- quickly, that even prolonged exposure (e.g., 20 tual learning and recovery. Systematic investiga- sessions) did not effect changes, and that changes tion of the relative contribution of training princi- observed for SGT were of similar rapidity. These ples in the treatment package would also help to observations support the conclusion that improve- identify the critically important elements associ- ments in the SGT phase were due to a generaliza- ated with successful learning and transfer. Finally, tion training effect, rather than due to additional the ability to modify sensory abilities experimen- dosage of stimulus exposure, or SST. tally opens up an experimental paradigm for future investigations of the relationship between sensa- Implications for Therapeutic tion and other abilities, such as pinch grip, and the Interventions and Further Investigations effect of improving sensation on motor function. Clinically significant improvements were ob- tained with training, providing justification for ACKNOWLEDGMENTS therapeutic intervention. Practice or exposure We thank the participants for their involve- alone was not usually sufficient to achieve the ment in the study and the therapists and manage- changes characteristic of perceptual learning. The ment at Royal Talbot Rehabilitation Centre, Cedar findings also suggest that the nature of training is Court Healthsouth Rehabilitation Hospital, and crucial to outcome. SST may be important if there Hampton Rehabilitation Hospital, Victoria, Aus- is a need to train specific work or daily living tasks. tralia, for their support. However, to facilitate more generalized improve- ments and minimize reliance on resource-inten- REFERENCES sive, task-specific training, a program that also 1. Schmidt RA, Lee TD: Motor Control and Learning: A Behav- includes variation within and across related train- ioral Emphasis. Champaign, IL, Human Kinetics, 1999 ing stimuli, intermittent feedback, and tuition of 2. Goldstone RL: Perceptual learning. Ann Rev Psychol 1998; 49:585–612 training principles is recommended. In the major- 3. Sathian K, Zangaladze A: Tactile learning is task specific but ity of cases, ten training sessions seem to be suffi- transfers between fingers. Percept Psychophys 1997;59: cient to achieve clinically significant improvement. 119–28 However, the goal of healthcare providers is not 4. Nagarajan SS, Blake DT, Wright BA, et al: Practice-related only to provide an effective training program but improvements in somatosensory interval discrimination are temporally specific but generalize across location, hemi- also an efficient training program. The highly spe- sphere and modality. J Neurosci 1998;18:1559–70 cific nature of training suggests that clinicians 5. Ahissar M, Hochstein S: Task difficulty and the specificity of need to accurately identify the specific sensory di- perceptual learning. Nature 1997;387:401–6 mension that requires training and use graded 6. Spengler F, Roberts TP, Poeppel D, et al: Learning transfer stimuli that cover the range of distinctive features and neuronal plasticity in humans trained in tactile discrim- for the perceptual dimension targeted. Programs ination. Neurosci Lett 1997;232:151–4 cannot rely on spontaneous transfer; they need to 7. Zihl J, Von Cramon D: Restitution of visual function in patients with cerebral blindness. J Neurol Neurosurg Psychi- systematically train for transfer. This issue of gen- atry 1979;42:312–22 eralization of training effects has importance in 8. Carey LM, Matyas TA, Oke LE: Sensory loss in stroke relation to rehabilitation of a range of abilities in patients: Effective tactile and proprioceptive discrimination stroke patients and to rehabilitation in general. training. Arch Phys Med Rehabil 1993;74:602–11 The present studies provide some guidance on 9. Yekutiel M, Guttman E: A controlled trial of the retraining of both the nature of transfer expected and on the the sensory function of the hand in stroke patients. J Neurol Neurosurg Psychiatry 1993;56:241–4 conditions of training required to achieve transfer. 10. Byl N, Roderick J, Mohamed O, et al: Effectiveness of sen- Further systematic investigation of these variables sory and motor rehabilitation of the upper limb following is indicated. For example, transfer across macro- the principles of neuroplasticity: Patients stable poststroke. spatial and microspatial features of surfaces will Neurorehabil Neural Repair 2003;17:176–91 help to determine if the boundary of spontaneous 11. Smania N, Montagnana B, Faccioli S, et al: Rehabilitation of somatic sensation and related deficit of motor control in transfer is at the level of a particular spatial pattern patients with pure sensory stroke. Arch Phys Med Rehabil or broader spatial features of roughness. Transfer 2003;84:1692–1702 across dimensions such as smoothness–roughness, 12. Carey LM: Somatosensory loss after stroke. Crit Rev Phys hardness–softness, and compressional elasticity Rehabil Med 1995;7:51–91 within multidimensional texture stimuli is also in- 13. Franklin RD, Allison DB, Gorman BS (eds): Design and dicated, particularly after SGT. Investigation of Analysis of Single-Case Research. Mahwah, NJ, Lawrence Erlbaum Associates, 1997 transfer across different joints of the upper limb for 14. Carey LM, Oke LE, Matyas TA: Impaired limb position sense proprioception or different fingers for texture is after stroke: A quantitative test for clinical use. Arch Phys also necessary. Identification of sites of cerebral Med Rehabil 1996;77:1271–8 reorganization underlying the recovery of somato- 15. Carey LM, Oke LE, Matyas TA: Impaired touch discrimina-

14 Carey and Matyas Am. J. Phys. Med. Rehabil. ● Vol. 84, No. 5 (Art: 162531 Input-mbb(v 4:48 05/11/3 10؍balt5/z79-phm/z79-phm/z7900505/z792796-05a browncm S

tion after stroke: A quantitative test. J Neurol Rehabil 23. Recanzone GH, Merzenich MM, Jenkins WM: Frequency 1997;11:219–32 discrimination training engaging a restricted skin surface results in an emergence of a cutaneous response zone in 16. Carey LM: Tactile and Proprioceptive Discrimination Loss cortical area 3a. J Neurophysiol 1992;67:1057–70 After Stroke: Training Effects and Quantitative Measure- ment [PhD thesis]. Melbourne, LaTrobe University, 1993 24. Karni A, Meyer G, Jezzard P, et al: Functional MRI evidence for adult motor cortex plasticity during motor skill learn- 17. Rosenthal R: Meta-analytic Procedures for Social Research, ing. Nature 1995;377:155–8 Revised Edition. Newbury Park, Sage Publications, 1991 25. Mountcastle VB: The columnar organization of the neocor- 18. Carey LM, Matyas TA, Oke LE: Quantitative measurement of tex. Brain 1997;120:701–22 tactile and proprioceptive discrimination in the upper limb after stroke. Stroke 2000;31:2816–7 26. Mesulam MM: From sensation to cognition. Brain 1998; 121:1013–52 19. Cormier SM, Hagman JD: Transfer of Learning: Contem- 27. Johnson KO, Hsiao SS: Neural mechanisms of tactual form porary Research and Applications. San Diego, Academic and texture perception. Ann Rev Neurosci 1992;15:227–50 Press, 1987 28. Clark FJ, Horch KW: Kinesthesia, in Boff KR, Kaufman L, 20. Directions for Using Semmes-Weinstein Monofilaments. Thomas JP (eds): Handbook of Perception and Human San Jose, North Coast Medical, 1987 Performance, chapter 13. New York, John Wiley, 1986 21. Annett M: A classification of hand preference by association 29. Gandevia SC, McCloskey DI, Burke D: Kinaesthetic signals analysis. Br J Psychol 1970;61:303–21 and muscle contraction. Trends Neurosci 1992;15:62–5 22. Roland PE: Somatosensory detection in patients with cir- 30. Carey LM, Abbott DF, Puce A, et al: Reemergence of acti- cumscribed lesions of the brain. Exp Brain Res 1987;66: vation with poststroke somatosensory recovery: A serial 303–17 fMRI case study. Neurology 2002;59:749–52

May 2005 Training of Somatosensory Discrimination 15