Individuals with autism spectrum disorder ( ASD ) show abnormalities during initial 1 and subsequent phases of

29 Sensorimotor impairments are common in ASD, but they are not well understood. 30 Here, we examined force control during initial pulses and the subsequent rise, sustained, 31 and relaxation phases of precision gripping in 34 individuals with ASD and 25 healthy 32 controls. Participants pressed on opposing load cells with their thumb and index finger 33 while receiving visual feedback regarding their performance. They completed 2 and 8 sec 34 trials during which they pressed at 15, 45 or 85% of their maximum force. Initial pulses 35 guided by feedforward control mechanisms, sustained force output controlled by visual 36 feedback processes, and force relaxation rates all were examined. Controls favored an 37 initial pulse strategy characterized by a rapid increase in and then relaxation of force 38 when the target force was low (Type 1). When the target force level or duration of trials 39 was increased, controls transitioned to a strategy in which they more gradually increased 40 their force, paused, and then increased their force again. Individuals with ASD showed a 41 more persistent bias towards the Type 1 strategy at higher force levels and during longer 42 trials and their initial force output was less accurate than that of controls. Patients showed 43 increased force variability compared to controls when attempting to sustain a constant 44 force level. During the relaxation phase, they showed reduced rates of force decrease. 45 These findings suggest that both feedforward and feedback motor control mechanisms 46 are compromised in ASD and these deficits may contribute to the dyspraxia and 47 sensorimotor abnormalities often seen in this disorder. 48 Introduction 49 Sensorimotor abnormalities are present in the majority of individuals with autism 50 spectrum disorder (ASD) (Baranek 1999; Fournier et al. 2010a). They emerge early in 51 infancy (Provost et al.2006; Bryson et al. 2007), and they appear to be familial (Mosconi 52 et al. 2010), but they have been studied far less frequently than the social-communication 53 and cognitive impairments that define the disorder. Multiple types of sensorimotor 54 impairments have been identified in ASD including reduced postural stability (Minshew 55 et al. 2007; Fournier et al. 2010b), atypical gait (Hallett et al. 1993; Vernazza-Martin et al. 56 2005), reduced coordination of upper limb movements (Mari et al. 2003; Glazebrook et al. 57 2007; Cook et al. 2013), macrographia (Fuentes et al. 2009) and atypical grasping 58 behaviors (David et al. 2009, 2012). But, the control processes and neurophysiological 59 mechanisms underlying these dysfunctions are not well understood. 60 Sensorimotor behavior involves integrating multiple distinct motor control 61 processes. Rapid movements are guided primarily by feedforward control mechanisms 62 that plan motor output faster than sensory feedback can be used to make online 63 adjustments (Prablanc and Martin 1992; Desmurget et al. 1999; Mari et al. 2003). Rapid 64 eye movements and control of initial manual motor output have been shown to be 65 impaired in ASD suggesting that feedforward processes may be compromised 66 (Glazebrook et al. 2007, 2009;David et al. 2009, 2012; Johnson et al. 2013; Mosconi et al. 67 2013; Schmitt et al. 2014). Sustained actions in which individuals attempt to maintain a 68 constant level of motor output rely more on sensory feedback mechanisms that allow 69 individuals to reactively adjust their motor behavior (Deutsch and Newell 2001, 2003). 70 Sustained eye movements have been shown to be less accurate in individuals with ASD 71 (Takarae et al. 2004), and sustained reaching movements show atypical kinematic 72 profiles (Mari et al. 2003; Glazebrook et al. 2007; Cook et al. 2013). Studies of motor 73 adaptation have indicated an increased reliance on proprioceptive relative to visual 74 feedback mechanisms in ASD suggesting sensorimotor feedback processes are disrupted 75 (Haswell et al. 2009; Izawa et al. 2012). 76 To assess feedforward and feedback motor control processes in ASD, we 77 examined initial and sustained force output during a test of visually guided precision 78 gripping. Precision gripping was studied because the ability to precisely regulate grip 79 forces is necessary for many activities of daily living known to be impaired in ASD, such 80 as writing and dressing (Fuentes et al. 2009). Precision gripping can be formulated as a 81 triphasic action involving initial increases in force output to grip an object, maintenance 82 of appropriate force levels to manipulate the object, and relaxation of grip forces to 83 release the object (Potter et al. 2006; Spraker et al. 2012). The rise phase in which 84 individuals increase their force to grip an object can be further decomposed into an initial 85 pulse and corrective pulses made in response to sensory feedback. Importantly, the initial 86 pulse is completed rapidly (~200-300 ms) and thus is believed to be controlled primarily 87 by feedforward processes. While prior studies have identified deficits in precision 88 gripping in ASD (David et al., 2009; 2012), systematic analyses of the distinct phases of 89 gripping behavior are needed to determine the motor control mechanisms that are 90 affected. 91 Examining the distinct phases of precision grip in ASD could provide important 92 insights into the disorder’s neural underpinnings. Multiple cortico-cerebellar circuits are 93 involved in generating internal action representations that consolidate feedforward 94 control processes (Miall 1998; Bastian 2006), integrating cortical and spinal sensory 95 afferents to modify outgoing motor commands (Stein and Glickstein 1992), and timing 96 agonist/antagonist muscle synergies during the release of force (Vilis and Hore 1980, 97 Serrien and Wiesendanger 1999). The cerebellum has been repeatedly implicated in 98 post-mortem and neuroimaging studies of ASD (Bailey et al. 1998; Bauman and Kemper 99 2005; Stanfield et al., 2008; Whitney et al. 2008), but the distinct circuits that are 100 disrupted and the impact of cerebellar pathology on sensorimotor behaviors in ASD have 101 not been established. 102 In the present study, we adapted a previously developed, objective approach to 103 differentiate distinct types of initial pulse control strategies during precision gripping 104 (Novak et al. 2000; Fishbach et al. 2005; Wisleder and Dounskaia 2007; Grafton and 105 Tunik 2011). We predicted that relative to controls, individuals with ASD would show 106 elevated rates of initial pulses characterized by rapid increases and then relaxation of 107 force rather than those characterized by more gradual increases in force. We also 108 expected reduced accuracy of initial pulses in ASD. Consistent with the hypothesis that 109 individuals with ASD show deficits in visual feedback control of motor output, we 110 predicted increased force variability during the sustained phase. Last, we hypothesized 111 lower rates of force relaxation in individuals with ASD suggesting a reduced ability to 112 rapidly terminate motor activity. 113 114 Method 115 Participants 116 Precision grip force was examined in 34 individuals with ASD and 25 healthy 117 controls between 5-15 years of age (Table 1). Participant groups were matched on age, 118 gender, handedness and nonverbal IQ 1 . Prior to testing, IQ was assessed using the 119 Wechsler Abbreviated Scale of Intelligence for individuals six years of age or older 120 (ASD=26; control=19). The Wechsler Preschool and Primary Scale of Intelligence 121 (ASD=4; control=4) or Differential Abilities Scales-II (ASD=1) were used for children 122 less than six years of age. 123 Individuals with ASD were recruited through community advertisements and the 124 clinical programs of the Center for Autism and Developmental Disorders at the 125 University of Texas Southwestern Medical Center. The diagnosis of ASD was established 126 using the Autism Diagnostic Inventory-Revised (ADI; Lord et al. 1994), the Autism 127 Diagnostic Observation Schedule – II (ADOS; Lord et al. 2012), and expert clinical 128 opinion based on DSM-V criteria. ASD participants were excluded if they had a known 129 genetic or metabolic disorder. Control participants were recruited from the community 130 and were required to have a score of 8 or lower on the Social Communication 131 Questionnaire (Berument et al. 1999). Control participants were excluded for current or 132 past psychiatric or neurological disorders, family history of ASD in first-, secondor 133 third-degree relatives, or a history in first-degree relatives of a developmental or learning 134 disorder, psychosis, or obsessive compulsive disorder. 135 No participants were taking medications known to affect motor function at the 136 time of testing, including antipsychotics, stimulants, or anticonvulsants (Reilly et al. 137 2008). All participants had corrected or uncorrected far visual acuity of at least 20/40. No 138 participant had a history of head injury, birth injury or seizure disorder. After a complete 139 description of the study, informed parental consent was obtained from parents or 140 caregivers, and children provided written assent. Study procedures were approved by the 141 local Institutional Review Board. 142 Apparatus and Procedures 143 Participants were seated in a darkened room 53 cm from the center of a 27-in. 144 computer screen (Fig. 1A). They were positioned on an adjustable chair so that visual 145 stimuli were presented at their eye level. Participants rested their forearm and elbow in a 146 relaxed position on a custom-made arm brace clamped to a table. Elbow position 147 remained stationary at 90o flexion throughout testing. Participants used their thumb and 148 index finger to press against two opposing ELFF-B4 precision load cells (Measurement 149 Specialties TM , Hampton, VA; 1.27 cm in diameter) secured to a custom grip device 150 attached to the arm brace. Analog signals from the load cells were amplified through a 151 Coulbourn (V72-25) resistive bridge strain amplifier. A 16 bit A/D converter was used to 152 sample the force output at 120 Hz. 153 Prior to testing, each participant’s maximum voluntary contraction (MVC) was 154 calculated for each hand. To determine each participant’s MVC, they were instructed to 155 press on the load cells with as much force as possible during three separate trials. The 156 mean of the maximum values for these trials was used as the estimate of each 157 participant’s MVC (Vaillancourt et al. 2003). During the precision grip test, participants 158 viewed two horizontal bars: a red/green target bar and a white force bar. The white force 159 bar moved upwards with increased force, and participants were instructed to press on the 160 load cells as quickly as possible when the target bar turned green so that the force bar 161 reached the height of the target bar. They also were instructed to keep the force bar as 162 close to the target bar as possible until it turned red again, and then to release the load 163 cells as fast as possible. The target was set to 15, 45 or 85% of each participant’s MVC 164 and its position was fixed at the center of the monitor. The location of the force bar was 165 varied as a function of the target force level to maintain a constant visual gain of 166 2.97pixels/Newton (visual angle 0.34o/N) across conditions. Thus, the distance between 167 the target and force bars was greater for trials with larger target force levels. 168 Participants completed 2 and 8 sec trials of precision gripping. During the 2 sec 169 test, two blocks of five trials were presented for each hand at each force level. Each force 170 trial was 2 sec in duration and alternated with 2 sec rest periods. A 15 sec rest block was 171 provided after each block of trials. During the 8 sec test, participants completed two 172 blocks of three trials for each hand at each force level. Eight sec trials were followed by 8 173 sec rest periods, and each block was separated by 15 sec of rest. For both tests, the same 174 hand was never tested on consecutive blocks. The order of different force levels was 175 randomized across blocks. The order of the two experiments (2 and 8 sec) was randomly 176 assigned to each participant. Prior to each experiment, all participants successfully 177 completed practice trials at 30% of their MVC using their dominant hand to demonstrate 178 that they understood task instructions. All participants were able to complete these 179 practice trials. 180 Force Data Analysis 181 Trials were excluded from analyses if the onset of force preceded the start cue. 182 For each participant, only conditions with >2 valid trials were included in the final 183 analyses. The number of participants included in group comparisons varied across 184 conditions but was similar across groups for each analysis (ASD: N=28-34; Control: 185 N=20-25). 186 Each force trace was low-pass filtered via a double-pass 2 nd order Butterworth 187 filter with a cut-off of 15 Hz. To examine initial pulse characteristics (see below), the 1 st , 188 2 nd and 3 rd derivatives of the force data were calculated in Matlab. Then, these derivative 189 profiles were smoothed with the same filter using a 6 Hz cut-off due to the inflated noise 190 induced from the differentiation procedure. 191 Force data was analyzed using a custom algorithm in Matlab. The grip force onset 192 was defined as the time point at which the rate of force increase first exceeded 5% of the 193 peak rate of force increase and remained and remained above this level for at least 100 194 ms (Fig. 1B) (Grafton and Tunik 2011). For the 2 sec test, the offset of the rise phase was 195 identified at 1 sec following the peak rate of force increase, or at the stop cue if this 196 occurred first. For the 8 sec test, the end of the rise phase was marked when the rate of 197 force increase fell below 5% of the peak rate of force increase, and the force level was 198 within 90% to 110% of the mean force of the sustained phase. Different procedures were 199 used for the two tests because participants were not able to consistently establish a period 200 of sustained force during the 2 sec trials. 201 The peak rate of force increase, duration and accuracy of the rise phase each were 202 examined. Force accuracy for the rise phase was calculated using the following formula: 203 Facc_ rise = Frise -Ft arget Frise +Ft arget (1) 204 where Frise and Ftarget represent the force level at the rise phase offset and the target force 205 level, respectively. This formula yields a unitless estimate of force accuracy ranging from 206 -1 to 1. Responses in which the participant’s force output accurately reached the target at 207 the end of the rise phase yield a Facc_rise=0, whereas negative values reflect force 208 undershooting and positive values indicate that force production exceeded the target force 209 level. 210 We decomposed the force rise phase into initial and secondary corrective pulses 211 using a previously developed and automated scoring algorithm (Novak et al. 2000; 212 Fishbach et al. 2005; Wisleder and Dounskaia 2007; Grafton and Tunik 2011). This 213 algorithm objectively defines the endpoint of the initial pulse at the first zero-crossing in 214 the force derivative traces. Pulses are categorized into different types depending on 215 whether the earliest zero-crossing after the peak rate of force increase is identified in the 216 first, second or third derivative trace (Fig. 2). The following pulse types were examined: 217 Type 1 (pulse-release): Type 1 initial pulses were characterized by an increase in 218 and then rapid reduction in force. Given that the corrective pulse was the opposite 219 direction of the initial pulse, the Type 1 initial pulse offset was identified at the first 220 zero-crossing from (+) to (-) in the 1 st derivative of the force time series following the 221 peak rate of force increase. 222 Type 2 (pulse-reaccelerate): Type 2 initial pulses were characterized by an 223 increase in force followed by a pause and then secondary increases in force which did not 224 temporally overlap with the initial pulse. The offset of Type 2 initial pulses were marked 225 at the first zero-crossing from (-) to (+) in the 2 nd derivative of the force output following 226 the peak rate of force increase. 227 Type 3 (overlapping pulses): Type 3 pulses involved increases in force followed 228 by one or more corrective increases in force that overlapped temporally with the initial 229 pulse. The offset of the initial pulse was marked at the first zero-crossing from (+) to (-) 230 in the 3 rd derivative of the force output following the peak rate of force increase of the 231 initial pulse. 232 We compared the rates at which individuals with ASD and healthy controls 233 produced each type of initial pulse across force levels and across the 2 and 8 sec tests. Eq. 234 (1) was used to define the accuracy of initial pulses. The peak rate of force increase and 235 duration of each initial pulse also were examined. 236 Sustained contractions were examined only for the 8 sec test. The first and last 237 seconds of the force time series were removed to minimize the influence of rise and 238 relaxation phases of the force response on sustained force measurements as we have done 239 previously (Fig. 1B) (Vaillancourt et al. 2003). Trials in which participants did not 240 sustain contractions for >5 sec, or the force level returned to zero for >1 sec were 241 excluded. The mean and coefficient of variation (CoV) of the de-trended sustained force 242 time series were examined. The CoV was calculated by dividing the variability of the 243 force time series by the mean force output and thus was used to examine sustained force 244 variability while controlling for differences in mean force output between groups. To 245 examine individuals’ ability to rapidly terminate force, we examined participants’ peak 246 rate of force decrease during the relaxation phase by identifying the minimum value of 247 the 1 st derivative of the force trace following the stop cue. 248 Clinical Measures 249 The ADI is a semi-structured parent/caregiver interview used to rate the level of 250 abnormality for each of the core symptom domains of ASD, including social impairment, 251 communication impairment, and restricted, repetitive behaviors (Lord et al. 1994; Rutter 252 et al. 2012). The ADOS is a semi-structured assessment of play, social abilities, 253 communication skills, and imaginative use of materials performed with each individual 254 with ASD by an examiner trained to research reliability. For both the ADI and ADOS, 255 higher scores reflect more severe abnormality in a given domain. These tests were used to 256 establish a diagnosis of ASD in participants and to examine the relationship between grip 257 force alterations and clinical features of ASD. 258 Statistical analysis 259 No significant effects of hand or interactive effects of hand and group were found 260 (all p’s>.05). Therefore, right and left hand performances were averaged for each force 261 level of each test. A series of repeated measure ANOVAs were conducted to compare 262 groups on force performance across force levels during the rise, sustain and relaxation 263 phases. Separate analyses were conducted for the 2 and 8 sec tests. Significant 264 interactions were examined using Bonferroni post hoc tests at each force level. Because 265 multiple participants did not have a sufficient number of trials (>2) for each initial pulse 266 type for each force level to compare force characteristics (e.g., rate of force increase, 267 duration, accuracy), comparisons of initial pulse characteristics were performed using 268 separate ANOVAs. Pearson correlation coefficients were used to examine the 269 relationships between force variables found to be different between groups and age, IQ, 270 and clinical ratings of ASD based on the social affect total score of the ADOS and the 271 social, communication and repetitive behavior algorithm scores of the ADI. 272 Results 273 Fig 3 shows raw traces of participants’ force output at 15% MVC of the 2 sec and 274 8 sec tests for four healthy controls and four representative participants with ASD. 275 Participants were selected based on the representativeness of their performance relative to 276 the group findings, and to include a broad range of MVCs (20-213.3 N). As can be seen 277 in Fig. 3, individuals with ASD showed a tendency to overshoot the target during the 278 initial pulse and produce increased sustained force variability. 279 Individuals with ASD had lower MVCs than controls for both their right 280 (ASD=53.6 N, SE=3.7 N; Control =68.5 N, SE=4.2 N) and left (ASD =53.3 N, SE=3.7 N; 281 Control =63.5 N, SE=4.2 N) hands (group main effect: F1,128=9.89, p=0.00). The 282 difference in strength between groups did not differ across hands (group × hand 283 interaction: F1,128=0.35, p=0.56). 284 Initial pulse characteristics. 285 Fig. 4 shows that for the 2 sec test, controls utilized a Type 1 strategy more 286 frequently than other strategies at 15% MVC (F2,168=10.64, p=0.00; Type 1=0.49, Type 287 2=0.28, Type 3=0.23). They shifted to using the Type 2 strategy more frequently than 288 other pulse types at 45% (F2,168=13.38, p=0.00; Type 1=0.35, Type 2=0.46, Type 3=0.19) 289 and 85% MVC (F2,171=22.56, p=0.00; Type 1=0.20, Type 2=0.58, Type 3=0.22). 290 ASD participants favored a Type 1 strategy at 15% MVC (F2,168=20.41, p=0.00; 291 Type 1=0.53, Type 2=0.25, Type 3=0.23). But, unlike controls, they also favored the Type 292 1 strategy at 45% MVC (F2,168=11.03, p=0.00; Type 1=0.46, Type 2=0.33, Type 3=0.23). 293 They shifted to a Type 2 strategy at 85% MVC (F2,171=11.47, p=0.00; Type 1=0.33, Type 294 2=0.46, Type 3=0.20). Individuals with ASD showed higher rates of Type 1 pulses 295 compared to controls at 45% and 85% MVC (45%: F1,168=6.28;p=0.01; 85%: F1,171=4.80, 296 p=0.03) and lower rates of Type 2 pulses relative to healthy controls at these higher force 297 levels (45%: F1,168=6.37; p=0.01; 85%: F1,171=4.78, p=0.03). For all force levels, both 298 groups used the Type 3 strategy less frequently than the Type 1 or 2 strategies. 299 During the 8 sec test, controls showed similar rates of Type 1 and Type 2 pulses at 300 15% and 45% MVC, and both strategies were used more frequently than Type 3 pulses 301 (15% MVC: F2,147=3.63; p=0.03; Type 1=0.38, Type 2=0.34, Type 3=0.19; 45% MVC: 302 F2,156=6.10; p=0.03; Type 1=0.39, Type 2=0.34, Type 3=0.13). Thus, relative to the 2 sec 303 test, healthy controls showed a more equal distribution of Type 1 and 2 pulses at 15% 304 MVC suggesting that they altered their control strategy in response to the increase in the 305 duration of the task. At 85% MVC, they favored Type 2 relative to Type 1 and Type 3 306 pulses (F2,138=13.84; p=0.00; Type 1=0.22, Type 2=0.55, Type 3=0.19). 307 In contrast to healthy controls, ASD participants continued to show a bias towards 308 Type 1 pulses at 15% MVC during the 8 sec test (F2,147=12.00; p=0.00; Type 1=0.52, 309 Type 2=0.22, Type 3=0.25). They showed a relatively even distribution of Type 1 and 310 Type 2 pulses at 45% MVC (F2,156=9.7; p=0.00; Type 1=0.43, Type 2=0.33, Type 3=0.18) 311 and then favored Type 2 pulses at 85% MVC (F2,138=15.05; p=0.00; Type 1=0.23, Type 312 2=0.51, Type 3=0.15). Individuals with ASD showed higher rates of Type 1 pulses 313 compared to controls at 15% MVC (F1,147=5.19; p=0.02); the control strategies used at 314 higher force levels did not differ between groups (45% MVC: F1,156=2.04; p=0.16; 85% 315 MVC: F1,138=0.35; p=0.56). 316 Comparisons of the accuracy, rate of force increase, and duration of each initial 317 pulse type are reported for individuals with ASD and healthy controls in Table 2. There 318 were no group differences in initial pulse characteristics for the 2 sec test. During the 8 319 sec test, individuals with ASD showed increased initial pulse overshoot compared to 320 controls at 15% MVC when using Type 1 or Type 3 pulses. They showed a reduced rate 321 of force increase at 85% MVC compared to controls for all pulse types. At 45% MVC, 322 the duration of their Type 2 pulses was shorter than for controls, whereas the duration of 323 their Type 3 pulses was longer. 324 Rise phase. 325 At the end of the rise phase, participants’ accuracy decreased with increases in 326 target force level for both tests (Fig. 5; 2 sec: F1.70, 93.22=33.66, p=0.00; 8 sec: F1.36, 327 76.00=62.20; p=0.00). For the 2 sec test, individuals with ASD overshot the target at 15% 328 MVC whereas healthy controls were closer to the target and tended to undershoot (F1,55 329 =9.00, p=0.00). While individuals with ASD tended to be less accurate than healthy 330 controls across other force levels and durations, neither the overall group differences or 331 the group × force level interactions were significant (p’s >0.05). 332 Sustained phase. 333 As expected, participants showed increases in mean force as the target force level 334 was increased (F1.08, 60.73=252.58, p=0.00) (Fig. 6-top). Compared with controls, 335 individuals with ASD showed reduced mean force overall, and this reduction was more 336 severe at higher force levels (target force × group interaction: F1.08, 60.73=9.63, p=0.00; 337 15% MVC: F1,56=2.62, p=0.11, 45% MVC: F1,56=6.64, p=0.01; 85% MVC: F1,56=9.27, 338 p=0.00). We examined the ratio of mean force to each participant’s target force level to 339 determine whether lower mean force in ASD was due to their lower MVCs. The group × 340 force level interaction was significant due to reduced mean:target force levels for the 341 ASD compared to the control group at 85% MVC (F1.16, 64.83=4.19, p=0.04; 15% MVC: 342 F1, 56=2.48,p=0.12, Control=1.06 N, SE=0.06N, ASD=1.18N, SE=0.05N; 45% MVC: F1, 343 56=0.58, p=0.45,Control=0.96 N, SE=0.02N, ASD=0.94 N, SE=0.02N; 85% MVC: F1, 344 56=4.05, p=0.048, Control=0.84 N, SE=0.03N, ASD=0.77N, SE=0.02N). Individuals with 345 ASD showed increased CoV compared to controls across target force levels suggesting 346 that increases in sustained force variability in ASD were evident even after controlling for 347 the modest decreases in mean force seen in ASD (Fig. 6-bottom; F1, 56=6.97, p=0.01). 348 Relaxation phase. 349 Analyses of 2 sec trials indicated that participants relaxed force more rapidly 350 during the relaxation phase at larger force levels (target force main effect: F1.19, 351 65.45=190.44, p=0.00) (Fig. 7). Individuals with ASD showed reduced rates of force 352 decrease compared to controls across force levels (target force × group interaction: F1.19, 353 65.45=6.27, p=0.01), particularly at 45% (F1, 55=11.48, p=0.00) and 85% MVC (F1, 55=7.24, 354 p=0.009). Across participants, the rate of force decrease was greater at larger force levels 355 compared to lower force levels during the 8 sec test as well (F1.08, 56.16=203.83, p=0.00). 356 Individuals with ASD showed reduced rates of force decrease (i.e., they were slower to 357 relax force) compared to controls across all force levels (F1.08, 56.16=6.53, p=0.01) 358 Clinical Correlations 359 Grip force performance was not associated with full scale or nonverbal IQ for 360 ASD participants (p’s >.05). For healthy controls, increased rates of Type 1 pulses at 361 15% MVC during the 8 sec test were associated with higher full scale IQs (r=0.49, 362 p=0.02). 363 Increased age was associated with a reduction in the rate at which Type 1 initial 364 pulses were used for the 2 sec test at both 45% and 85% MVCs for individuals with ASD 365 only (r=-0.38, p=0.03). Age was not associated with the rate of different pulse types for 366 healthy controls. Both groups demonstrated age-related increases in mean sustained force 367 at both 45% and 85% MVC (ASD: r=0.70, p=0.00; control: r=0.80, p=0.00) and 368 reductions in sustained force CoV across all target force levels (ASD: r=-0.37, p=0.00; 369 control: r=-0.72, p=0.00). Age-related reductions in CoV were stronger for controls 370 compared to individuals with ASD (Fisher’s Z=1.86, p=0.03). Increased age also was 371 associated with increased rates of force decrease during the relaxation phase for all force 372 levels on the 2 sec (ASD: r=-0.70, p=0.00; control: r=-0.70, p=0.00) and 8 sec tests (ASD: 373 r=-0.70, p=0.00; control: r=-0.70, p=0.00). The strength of age-associated decreases in 374 rate of force relaxation was not different between controls and individuals with ASD 375 (p>.05). 376 Increased rates of Type 1 pulses at 45% and 85% MVC of the 2 sec test were 377 associated with more severe clinically rated ADI-R social-communication abnormalities 378 in ASD (r=0.42, p=0.02). No other relationships between force performance and clinical 379 ratings of ASD symptoms were significant. 380 381 Discussion 382 In the present study of precision gripping, we found that individuals with ASD 383 utilize an initial pulse strategy characterized by rapid increases in and then release of 384 force more frequently than controls. While controls also use this strategy when target 385 force levels and grip durations are relatively low, they adapt to increased demands on 386 their force output by transitioning to a strategy in which they increase force more 387 gradually, pause, and then increase their force output again. Our results also indicate that, 388 when sustaining a constant force level, individuals with ASD show increased output 389 variability suggesting that they have a reduced ability to translate visual feedback 390 information into precise motor commands. Last, patients showed a consistent reduction in 391 the rate at which they released their grip indicating a reduced ability to rapidly terminate 392 force output. 393 Precision grip abnormalities in ASD 394 While prior studies have suggested that individuals with ASD show a reduced 395 ability to integrate load and lifting forces during gripping (David et al. 2009, 2012), ours 396 is the first known study to identify differences in the underlying strategy used by 397 individuals with ASD to control initial force output. The duration of initial pulses ranged 398 between 200-300 ms suggesting that they are completed before visual feedback is likely 399 to have a large impact on force output and thus are largely controlled by feedforward 400 mechanisms (Miall et al. 1998; Kawato 1999; Wisleder and Dounskaia 2007). Our 401 findings that individuals with ASD utilize an atypical initial pulse strategy, and that the 402 accuracy, rate of force increase and duration of initial pulses are abnormal in ASD 403 suggest that feedforward mechanisms involved in controlling initial motor output are 404 disrupted. 405 From a sensorimotor efficiency perspective, healthy controls’ transition from a 406 Type 1 to a Type 2 initial pulse strategy at higher force levels and prior to longer 407 sustained contractions is advantageous for reducing the operative cost on the 408 neuromuscular system. While increasing force output involves temporal coordination of 409 the agonist muscles of the thumb and index finger (Nowak et al. 2004; Bastian 2006; 410 Potter et al. 2006), decreasing grip force requires coordination of both agonist and 411 antagonist muscles of the fingers and hand (Day et al. 1998; Bastian 2006; Potter et al. 412 2006). The reduced mechanical requirements of a Type 2 approach likely explains why 413 individuals tend to produce lower force levels than required when first manipulating an 414 object of unknown weight as this approach allows them to adjust their force output more 415 efficiently (Nowak et al. 2004). In addition, producing excessive force during initial 416 contractions and then relaxing force levels not only increases the difficulty of the action 417 as more muscles are involved, but it also may lead to muscular fatigue and, therefore, 418 disrupt control of subsequent force output. 419 Because our procedure for differentiating primary pulses was based on derivatives 420 of force output that effectively amplify noise, it is possible that estimates of the timing 421 and number of zero-crossings in third derivative traces may reflect Type 3 pulse 422 strategies as well as low amplitude oscillations from mechanical (e.g., electrical noise) 423 and biological sources. A conservative filter cut-off (6 Hz) was used to minimize artifact 424 from higher frequency mechanical noise in derivative traces, but oscillations reflecting 425 peripheral (e.g., joint, muscle fiber, motor unit, motor neurons, etc.) and central nervous 426 system processes are more difficult to distinguish from pulse strategies. These 427 oscillations could contribute to overestimates of Type 3 pulses and underestimates of 428 Type 3 pulse durations. While we found that Type 3 pulses were less common than Type 429 1 and 2 pulses and of similar duration (see Table 2), Type 3 pulse results should be 430 interpreted with caution due to the possible influence of both electrical and biological 431 noise on our differentiation procedure. 432 Individuals with ASD also showed reduced mean force and increased force 433 variability relative to controls when attempting to sustain a constant level of force. 434 Reductions in mean force were largely reflective of decreased strength as indicated by 435 patients’ lower MVCs (Fellows et al. 2001; Hardan et al. 2003; Kern et al. 2011). But, 436 after adjusting for overall decreases in mean force, individuals with ASD still showed 437 increased force variability suggesting they have a reduced ability to accurately adjust 438 their motor output online (Gepner and Mestre 2002). This impairment may represent a 439 major component of the difficulties in performing skilled tasks of the hands and fingers 440 that often are seen in individuals with ASD (Dziuk et al. 2007; Fuentes et al. 2009). 441 During the relaxation phase, patients showed reduced rates of force release. While 442 individuals with ASD appear to show a protracted course of movement deceleration when 443 making rapid saccadic eye movements (Glazebrook et al., 2009), to our knowledge, their 444 ability to terminate force output has not been previously examined. These results provide 445 behavioral evidence suggesting that the termination of grip is impaired in ASD, which 446 could be related to abnormal agonist and antagonist muscles (Vilis and Hore 1980). 447 Direct electromyographic (EMG) measurements of muscle activation patterns during 448 precision gripping and releasing in ASD are needed to determine the mechanisms 449 underlying patients’ reduced rates for force release. 450 Neural mechanisms underlying visuomotor abnormalities in ASD 451 The profile of visuomotor alterations seen here in ASD implicates dysfunctions in 452 cortico-cerebellar networks involved in visuomotor control. The cerebellum appears to be 453 a particularly important region both for generating internal action representations used to 454 predictively control initial motor output (Kawato 1999; Bastian, 2006) and translating 455 visual feedback information from posterior parietal cortex into reactive motor 456 adjustments to control sustained motor behaviors (Coombes et al. 2010; Stein and 457 Glickstein 1992; Vaillancourt et al. 2003). Prior studies of patients with cerebellar lesions 458 have documented patterns of deficit during precision gripping that are similar to those we 459 observed here in ASD, including excess initial force output, increased sustained force 460 variability, and decreased rates of force relaxation (Mai et al. 1988; Müller and Dichgans 461 1994; Serrien and Wiesendanger 1999; Fellows et al. 2001; Nowak et al. 2002, 2004). 462 The hypothesis that cerebellar alterations may contribute to precision gripping 463 abnormalities in ASD also is supported by numerous post-mortem studies of ASD 464 patients documenting Purkinje cell and deep nuclear pathology (Bailey et al. 1998; 465 Bauman and Kemper 2005; Whitney et al. 2008), and neuroimaging studies showing both 466 structural and functional abnormalities of cerebellar circuits in ASD (Courchesne et al. 467 1988; Allen and Courchesne, 2003; Catani et al. 2008; Mostofsky et al. 2009; Groen et al. 468 2011). Thus, our neurobehavioral findings suggest that cerebellar pathology may 469 contribute to the visuomotor deficits characteristic of this disorder. 470 While the profile of precision grip abnormalities seen here in ASD implicates the 471 cerebellum, it also is possible that alterations of other cortical and subcortical regions 472 contribute to visuomotor deficits in this disorder (Valvano and Newell 1998; Gordon and 473 Duff 1999; Hermsdörfer et al. 2003, 2004; Quaney et al. 2005). During precision 474 gripping, parietal and motor cortices are involved in processing sensory feedback and 475 generating motor commands, respectively (Vaillancourt et al. 2006). Patients with focal 476 lesions of premotor, primary motor and parietal cortices have been shown to demonstrate 477 excessive initial grip force and increased sustained force variability (Mostofsky et al. 478 2009; Eidenmüller et al. 2014). Patients with Parkinson’s disease also have been found to 479 demonstrate increased force variability during precision gripping implicating the basal 480 ganglia (Neely et al. 2013). While neuroimaging studies are necessary for establishing the 481 neural mechanisms underlying visuomotor impairments in ASD, our findings collectively 482 suggest that cerebellar pathology and dysfunctions in cortical and striatal circuits may 483 contribute to the neurodevelopmental alterations seen in this disorder. 484 Clinical associations 485 While both individuals with ASD and healthy controls showed age-related 486 reductions in sustained force variability, this improvement was stronger in controls 487 suggesting that visuomotor deficits in ASD may persist during later childhood and into 488 adolescence. Prior studies of healthy development have indicated that age-related 489 decreases in force variability reflect an increased ability to utilize visual and haptic 490 information to refine ongoing performance (Deutsch and Newell 2001, 2003; Potter et al. 491 2006). Our findings indicate that individuals with ASD show dysmaturation of these 492 sensory feedback processes that are persistent and thus may be important targets for 493 interventions throughout development. 494 The association between increased rates of Type 1 initial pulses and more severe 495 social-communication abnormalities in ASD suggests that these deficits may reflect a 496 common neurodevelopmental mechanism, such as cerebellar dysfunctions (Wang et al. 497 2014). In addition to being critical to sensorimotor control, the cerebellum also has been 498 shown to be involved in social and cognitive development (Stoodley and Schmahmann 499 2009). Further, decreased cerebellar volume has been found to be associated with 500 increases in the volume of prefrontal cortices involved in cognitive and social processes 501 in ASD (Carper and Courchesne 2005). 502 It also is possible that visuomotor deficits contribute to the development of 503 social-communication impairments in affected individuals. Evidence for this hypothesis 504 comes from findings that early sensorimotor abnormalities in ASD are associated with 505 more severe social-communication features later in life (Sutera et al. 2007), and that 506 sensorimotor developments in infancy and toddlerhood are important for increasing the 507 quality and frequency of social interactions and language learning opportunities (Gallese 508 et al. 2013; LeBarton and Iverson 2013). Visuomotor impairments were not associated 509 with IQ or with the severity of repetitive behaviors indicating that they may be selectively 510 associated with social-communication dysmaturation in ASD. 511 Summary 512 Our results demonstrate that individuals with ASD show visuomotor impairments 513 affecting initial motor output, sustained force output, and the ability to rapidly release 514 force. Further, we provide novel evidence that reduced control of initial pulses in ASD 515 may reflect an atypical control strategy in patients, and a failure to flexibly adapt 516 feedforward control strategies to changing task demands. Our finding that alterations in 517 feedforward control strategies are associated with more severe social-communication 518 abnormalities suggests that the sensorimotor impairments present in the majority of 519 individuals with ASD may play a more central role in this disorder than previously 520