You are on page 1of 48

1 Title page

2 Title:

3 Strategies to develop putative biomarkers to characterize the female phenotype with


4 Autism Spectrum Disorders

5 Elizabeth B. Torres1,2, Robert W. Isenhower1, Polina Yanovich3, Gwendolyn Rehrig1, Kimberly


6 Stigler4, John Nurnberger4, Jorge V. José5,

7 1 Rutgers University – Psychology Dept.


8
9 2 Rutgers University –Rutgers Center for Cognitive Science, Center for Computational
10 Biomedicine Imaging and Modeling (Computer Science), Movement Disorders, Neurology,
11 Indiana University Medical School Indianapolis.
12
13 3 Rutgers University – Computer Science Department
14
15 4 Indiana University – Dept. of Psychiatry, Indiana University School of Medicine, Christian
16 Sarkine Autism Treatment Center, Riley Hospital for Children, Indianapolis
17
18 5 Indiana University – Department of Physics, Department of Cellular & Integrative
19 Physiology, Medical School, Indiana University.
20
21 Corresponding Author: Elizabeth B Torres

22 Rutgers University, Psychology Department

23 152 Frelinghuysen Rd.

24 Piscataway, NJ, 08854

25 (848) 445-8909 and ebtorres@rci.rutgers.edu

26 Running Head: Characterizing the Female with ASD

27 Key words – sex differences, autism, kinaesthetic input, decision-making, movement


28 decision time, stochastic signatures

29 Financial Disclosure/Conflict of Interest: There is no conflict of interest from any of the


30 authors.
1
31 Abstract
32 Current observational inventories used to diagnose Autism Spectrum Disorders (ASD) apply
33 similar criteria to females and males alike, despite developmental differences between the
34 sexes. Recent work investigating the chronology of diagnosis in ASD has raised the concern
35 that females run the risk of receiving a delayed diagnosis, potentially missing a window of
36 opportunity for early intervention. Here we retake this issue in the context of the objective
37 measurements of natural behaviors that involve decision making processes. Within this
38 context, we quantified movement variability in typically developing (TD) individuals and
39 those diagnosed with ASD, across different ages. We extracted the latencies of the decision
40 movements and velocity-dependent parameters as the hand movements unfolded for two
41 movement segments within the reach: movements intended towards the target and
42 withdrawing movements that spontaneously, without instruction, occurred incidentally. The
43 stochastic signatures of the movement decision latencies and the percent of time to
44 maximum speed differed between males and females with ASD. This feature was also
45 observed in the empirically estimated probability distributions of the maximum speed
46 values, independent of limb size. Females with ASD showed different dispersion than males
47 with ASD. The distinctions found for females with ASD were better appreciated when
48 compared to those of TD females. In light of these results, behavioral assessment of autistic
49 traits in females should be performed relative to TD females to increase the chance of
50 detection.
51

2
52 Introduction
53
54 The current diagnosis of Autism Spectrum Disorder (ASD) (American Psychiatric
55 Association 2000) does not make provisions to distinguish between males and females
56 despite known developmental sex differences across a broad range of parameters, from
57 physiological to cognitive milestones, including language. Typically developing boys are
58 known to lag behind girls initially, but eventually boys catch up (Stromswold 1998) and, by
59 adulthood, some complementary sex differences may emerge (Adam et al. 1999; Der and
60 Deary 2006; Kaushanskaya et al. 2011), suggesting parallel learning paths.
61
62 In children who have already been diagnosed with ASD, girls—particularly those
63 without marked cognitive impairments—may have been formally identified at a later age
64 than boys, which may have impeded referral for early interventions (Giarelli et al. 2010).
65 Because the diagnosis of ASD is centered on cognitive impairments (ADOS-(Berument et al.
66 1999); GARS-2- (Campbell 2005; Lecavalier 2005)) and the observation that there is a large
67 sex difference in the prevalence of ASD—approximately 5:1 boys to girls (Lord and Bishop
68 2010; Mandy et al. 2011; Newschaffer et al. 2007; Volkmar et al. 1993), it seems important
69 to develop non-invasive tests on cognitive performance (e.g., decision-making tasks) that
70 may be easily administered in a clinical setting to identify possible systematic sex
71 differences across ages. This information would be extremely valuable given its implications
72 for the correct standardization of sex-based cognitive metrics. Furthermore, sex differences
73 in brain development have been used to anchor cognitive theories of ASD (Baron-Cohen
74 2002; Baron-Cohen et al. 2003) with support from subjectively self-reported surveys,
75 inventories, and meta-analyses (Lai et al. 2011). At present, however, no operational
76 definition has been provided that permits fast, automatic, and objective screening for sex
77 differentiation in clinical settings. This is particularly so as the individual ages and his/her
78 sensory-motor systems adapt to new developmental changes.
79
80 The traditional approaches to ASD research across all areas of cognition, perception and
81 action have focused thus far on disruptions of goal-directed and intentional aspects of
82 behavior, assumed to be centrally driven. These include planning, coordination and motor
83 learning (Fournier et al. 2010a; Fournier et al. 2010b; Gidley Larson et al. 2008; Gowen et al.
84 2008; Haswell et al. 2009; Izawa et al. 2012; Jansiewicz et al. 2006; Jones and Prior 1985;
85 Minshew et al. 2004; Mostofsky et al. 2006; Noterdaeme et al. 2002; Rinehart et al. 2001;
86 Rogers et al. 1996; Williams et al. 2001) just to mention a few.
3
87 Behavior however, flows continuously rather than as a discrete set of goal-directed
88 segments. Although there is intentionality in much of what we do, the fluidity of our actions
89 heavily depends on spontaneous segments “gluing” our goal-directed acts (Torres et al.
90 2011). Very little is known about the potential contributions of reflexive, spontaneous and
91 automatic reactions that without explicit awareness co-occur in complex behaviors. Some of
92 these types of motions are known to be disrupted in infants that go on to receive a diagnosis
93 of ASD (Teitelbaum et al. 1998) and persist in excess (Minderaa et al. 1985; Reed 2007) or
94 are different in older children and adolescents with ASD (Torres 2012; Torres et al. 2013a).
95 Spontaneous aspects of motor control also breakdown when sub-cortical structures are
96 compromised in adults (Torres et al. 2011) and the endogenous feedback from the rotating
97 arm-joints breaks down. A large body of work from motor control (Redgrave and Coizet
98 2007; Redgrave et al. 2010; Shadmehr and Wise 2005; Smith et al. 2011) suggests that
99 peripheral motor input and sub-cortical structures are important to develop a good balance
100 between intended and automatic aspects of behavior. However, the body of work of
101 movement neuroscience has yet to make contact with the field of autism. The research
102 literature in autism focuses at present on cognitive and social issues, as if these were
103 disconnected from movement and movement sensing.
104
105 In autism, a disorder diagnosed after 3 years of age, the links and pathways between
106 voluntary and automatic structures for central- and peripheral-based control may have
107 developed differently. Problems with movement and movement sensing have been reported
108 in ASD (Donnellan et al. 2012; Donnellan and Leary 2012; Savarese 2013). These problems
109 could be objectively quantified at the motor output flow to begin elucidating the various
110 contributions of different aspects of motor planning and execution to these disturbances. We
111 know that several of the cortical and sub-cortical structures that are targeted by General
112 Somatic Afferent (GSA) fibers are reportedly impaired in ASD along with anomalies
113 involving central and peripheral synapses (Amaral and Corbett 2003; Amaral et al. 2008;
114 Breece et al. 2012; Damasio and Maurer 1978; Jacobson et al. 1988; Maurer and Damasio
115 1982; 1979; Mostofsky et al. 2009; Nordahl et al. 2012; Qiu et al. 2010; Rinehart et al. 2002;
116 Schumann et al. 2004; Takarae et al. 2007). Nevertheless, features of movement kinematics
117 from the peripheral limbs are not well understood in ASD. It is not known if sex differences
118 could be found in relation to such parameters.
119
120 Disturbances in the peripheral nervous systems (PNS) related to self-regulation, body
121 autonomy and volitional control also abound in ASD. Some of these may impact movement
122 and could be reflected in the continuous flow of motions, including the spontaneous
4
123 segments of natural behaviors. Some of the well known peripheral issues are problems with
124 the autonomic nervous system (ANS), involving the enteric (gastro-intestinal) subsystems of
125 the PNS (Ashwood et al. 2003; Buie et al. 2010; de Magistris et al. 2010; Kushak et al. 2011;
126 MacFabe et al. 2011; Mazurek et al. 2013; Molloy and Manning-Courtney 2003); problems
127 with sleep and the circadian rhythms are also common (Bourgeron 2007; Glickman 2010);
128 unusual and unpredictable pain and temperature dysregulation are well documented,
129 particularly in autism of known etiology (Bandstra et al. 2012; Dubois et al. 2010; Klintwall
130 et al. 2011; Nader et al. 2004; Tordjman et al. 2009; Zeidan-Chulia et al. 2011); etc. These are
131 however often downplayed and interpreted as “co-morbid”, secondary symptoms in the face
132 of more obvious differences in the development of social and communicative skills. It will
133 remain challenging to objectively screen and treat ASD based on observation and verbal
134 reports of behaviors, particularly when the generation and control of social behaviors
135 involves all aspects of motor control: planning, representation, sensory-motor feedback,
136 automatic body orienting, self-regulation, body self-autonomy, spontaneous variability, etc.
137
138 The disembodied, static and centrally-driven view of autism has been recently
139 challenged by a series of studies involving various motor acts reported in over 30 peer-
140 reviewed articles in a research topic of the open access platform, Frontiers in Neuroscience
141 (http://www.frontiersin.org/integrative_neuroscience/researchtopics/autism_the_movement_persp
142 ectiv/801). In particular, studies of motor variability in ASD posed as a stochastic process
143 over time (Torres 2012; Torres et al. 2013a; Torres et al. 2013d) have found that the various
144 signatures of speed-dependent variability appear to undergo a maturation process governed
145 by a scaling power law (Torres et al. 2013a). As we typically age, the stochastic signatures of
146 velocity-dependent parameters during pointing motions shift with age from random and
147 noisy to predictable and reliable. However in ASD, such transition is absent. Regardless of
148 age, verbal capabilities or gender, all individuals with ASD in this recent study remained
149 within the statistical ranges of the typically developing 3-4 year old participants. The
150 participants with ASD did not experience the age-dependent shifts in stochastic motor
151 signatures that the motor patterns of typically developing folks manifested over time. The
152 predictable and verifiable patterns of motor variability in our motions typically enable
153 adequate anticipation of impending hand velocities from prior actions and can be
154 informative of levels of intent during motor learning (Torres 2013a; b).
155
156 The recent findings in ASD invite to think about predictable and reliable motor
157 variability not only as a form of motor output, but also as a form of re-afferent feedback
158 necessary for the acquisition and maintenance of a stable movement percept, anticipatory
5
159 planning and volitional control. What features(s) of this bundle of peripheral limb
160 kinematics information may be of use to the central nervous system in the modulation and
161 control of our actions remains an open question. However, previous evidence from motor
162 illusions induced by muscle tendon vibrations have suggested that the perception of the
163 velocity of illusory limb movements depends on the frequency and amplitude of the applied
164 mechanical stimulation (Clark et al. 1979; Roll and Vedel 1982; Vedel and Roll 1982) and
165 that such systematic covariations may indicate that muscle receptors, such as the dynamics-
166 sensitive primary spindle endings, may be able to code movement velocity (Matthews 1982;
167 Matthews 1981).
168
169 Velocity is a rich kinematic parameter that contains spatial information in its directional
170 component and temporal information in its distance-metric dependent magnitude. These
171 also appear to be encoded by neurons along the sensory-motor pathways of the CNS. For
172 example, during reaching movements, several features of velocity are encoded by the
173 primary motor cortex (Georgopoulos et al. 1986; Redish and Touretzky 1994; Schwartz
174 1994). But these are also planned ahead by neurons in the reach regions of the posterior
175 parietal cortex, as animals learn complex curved trajectories in real time (Torres et al.
176 2013c). In these areas velocity-related ascending afferent inputs have also been reported
177 (Prevosto et al. 2010; 2009; 2011).
178
179 At the behavioral level, features related to movement velocity are often studied as
180 averaged quantities taken over many repetitions (Shadmehr and Wise 2005). Yet the
181 frequency of fluctuations of velocity-dependent parameters continuously accumulated over
182 time, as we repeat our actions, may be a form of natural vibration that the muscular and joint
183 afferent fibers could process, help interpret and send as feedback to the control areas of the
184 CNS. Several of these central areas are thought to form and anticipate a reliable motor
185 percept, (Hauschild et al. 2012; Kawato and Wolpert 1998; Mulliken et al. 2008; Wolpert et
186 al. 1998) even in the absence of active efferent commands from ongoing movements (Torres
187 et al. 2013c). In particular in the parietal areas, the separation between sensory input and
188 abstract planning has been demonstrated across several contexts (Andersen and Buneo
189 2002; Hwang et al. 2013; Torres et al. 2013c). Based on this prior research we examine here
190 velocity-dependent parameters from the continuous flow of natural, unconstrained hand
191 motions.
192
193 We hypothesize that differences in velocity-depending variability (taken as a bundle
194 of kinesthetic re-afferent inputs and motor outputs) may exist between ASD and
6
195 controls, and may be critical to further differentiate sex-based manifestations of the
196 disorder. This hypothesis is further motivated by reported sex differences in the physiology
197 relevant to sensory feedback from the PNS , including differences in pain-related input
198 (Derbyshire et al. 2002; Hashmi and Davis 2009; Racine et al. 2012), and reported sex
199 differences in smooth muscle vascular tone (Thompson and Khalil 2003), as well as sex
200 differences in muscle fibre size and type distribution in thoracic and lumbar regions
201 (Mannion et al. 1997).
202
203 In the present paper we focus on the values and timing of the peak velocity during
204 upper limb pointing motions subject to different levels of decision-making under
205 manipulations of the cognitive load in the visual stimuli for action. We ask if the timing
206 information tied to the continuous flow of movements could reveal some sex differences in
207 typical controls, between deliberate and spontaneous segments of the motor decisions. We
208 further ask whether such differences would serve to separate in any way males and females
209 with a diagnosis of ASD.

210 Materials and Methods

211 Participants
212 78 participants (34 ASD and 44 TD) ranging from 3.5 to 61 years of age with varying
213 intellectual capabilities were examined. Within the 44 TD participants, 22 were college
214 educated, ranging from students to professors. They performed a simple biomechanical
215 version of the pointing task (pointing to a dot) where no target dependent decisions were
216 included. The rest performed the pointing task and a match-to-sample version of the task
217 where several layers of decision making were required according to the cognitive loads
218 associated to the targets. The schematics of the the set up are shown in the Appendix.
219 Individuals with ASD had reported IQ ranges between 40 and 110. For the typically
220 developing (TD) individuals, IQ was reported as 90 and above, with education spanning
221 from pre-school to college levels. Demographic information is listed in Tables 1-2.
222
223 Parents signed parental consent for the children, adolescents provided assent, and
224 adults provided their consent. The protocol was approved by the Institutional Review
225 Boards of Rutgers University and Indiana University in compliance with the Declaration of
226 Helsinki. For participants with ASD, Tables 1 and 2 report the scores from various pencil-
227 and-paper diagnostic inventories. Licensed clinicians administered the Stanford-Binet 5th
228 Edition intelligence, the Autism Diagnosis Observation Scale (Lord et al. 2000)

7
229 (communication + social scores as well as the Gilliam Autism Rating Scale 2nd Edition
230 (Gilliam 2006), which confirmed diagnosis where appropriate (e.g. for non-verbal teenage
231 girls for whom none of the four ADOS modules currently available are appropriate).

232 Task and Apparatus


233 Subjects sat comfortably in front of a computer screen. The task consisted of natural
234 pointing motions towards a target on the computer screen (Appendix Figure.) All motions
235 included the forward reach intended to touch the screen monitor, and also the intermediate
236 segments spontaneously performed, beneath the subject’s awareness. These segments were
237 incidental to the main goals of the task and supplemented the goal-directed segments but
238 were not instructed in any way.
239
240 In the match to sample version of the pointing task, targets appeared on a touch
241 screen monitor (DELL with 3ms delay) when subjects touched the bottom center location of
242 the monitor. The touch evoked the sample stimulus to match with one of two possible
243 targets on the upper corners of the monitor. Upon landing the hand at the target of choice,
244 the next trial was initiated. The participants controlled the length of the trial and the flow of
245 the experiment. A motion caption system (Polhemus Liberty, 240Hz) recorded the
246 movements concurrently with the screen touches and the visual stimulus presentation, all
247 synchronized to the same CPU. A computer interface logged and time-stamped the
248 presentation of the stimulus (Freeman and Ambady 2010). The software developed in house
249 collected the screen touches. In the simpler pointing task, subjects had to touch a dot on the
250 screen. They then brought the hand to rest, but as in the match to sample pointing version,
251 this segment was not instructed.
252
253 The continuous flow of motion was harnessed, rather than only registering the goal-
254 directed trial by trial segments. The Fig.1A shows sample hand trajectories from this
255 continuous flow. The hand positional trajectories were obtained and the touches and
256 physical plane of the monitor were used as landmarks of target-contacting segments.
257 Backtracking along the speed temporal profile from the touch-pause (contacting the screen
258 plane) to the previous pause yielded goal-directed segments. The other spontaneously
259 withdrawing motions were also continuously harnessed (from the touch to the next pause,
260 as the distance from the touch screen plane grew in value.) Two sample goal-directed
261 trajectories are shown in black in Fig.1A towards the green targets on the screen plane and
262 two blue spontaneous retracting trajectories are also sown. The stars mark the peak

8
263 velocities and the time in milliseconds when they happened. The circles show the pauses
264 along the trajectories.
265
266 First-order changes in position over time (velocity) were obtained (Fig.1B) using the
267 smoothing and derivative functions from the Spline toolbox in MATLAB (MATLAB version
268 2012a, Natick, Massachusetts, TheMathWorks Inc.) For each velocity trajectory, the
269 instantaneous length of the three dimensional velocity vectors was obtained using the
270 Euclidean norm. A speed profile as a function of time was continuously obtained. The values
271 of speed maxima were computed and 5% of the overall maximum speed was used as a cutoff
272 to separate pauses from movement and to delineate movement segments. Specifically, the
273 speed minima (circles in Fig.1) were used as criteria (along with the touches and the
274 physical plane of the screen) to separate the goal directed motions from all the other
275 incidental ones. Further details about the set up and methods are explained in (Torres et al.
276 2013a). Figures 1C and 1D zoom in the two speed profiles from the trajectories in 1A
277 superimposed and aligned to the 5% cutoff.
278
279 For each participant, thousands of such motion segments were continuously
280 collected as the participants interacted with the touch screen. The fluctuations of the peak
281 velocity and of the time to reach this peak value were accumulated to build frequency
282 histograms and for each individual experimentally estimate the underlying probability
283 distributions governing such velocity-dependent parameters over time. The time-scale of a
284 session was 15-20 minutes. For example in Figure 1B there are 40,000 frames which at the
285 sampling rate of 240Hz give us 16.66 seconds worth of continuous motions. In a 20 min
286 session we accumulate hundreds of values of the peak velocity and of the time to reach these
287 maxima to use in the experimental estimations of the probability distributions. In this sense
288 we do not need to perform a fixed number of trials per person, as we are not averaging the
289 values of such parameters across trials. We are not a priori hand-picking homogeneous
290 groups to perform significant hypothesis testing. We are rather estimating the probability
291 distributions of each individual, their parameters and the shifts in such parameters as a
292 function of stimuli, decision load, etc. and letting the data reveal self-emerging clusters from
293 the inherent patterns of variability.

294 Kinematics parameters of interest


295 Sensory-motor-related parameters included the peak velocity (m/s) and the time (s) to attain
296 this value in each segment defined from the trough to the peak of each speed profile. We
297 normalized the peak velocity to control for differences in limb size (used the peak velocity

9
298 divided by the sum of the peak velocity and the averaged trial speed) (Lleonart et al. 2000;
299 Mosimann 1970). We also considered differences in the duration of the segments and
300 obtained the percent of time to reach the speed maximum relative to the total time of the
301 movement segment.
302
303 Decision-Making parameters included the accuracy of the decision in the match-to-sample
304 task (measured as the percent correct) and the movement decision latency (s), denoted
305 decision time, DT. To obtain the latency parameter, we measured the total time (s) from the
306 onset of the stimulus (evoked by the participant’s screen touch) to the target touch.
307 The movement decision time includes the reaction time, the time spent deciding, and the
308 actual movement time. Subtracting the movement time (given by the speed profile between
309 the two relevant minima) provides the latency of the movement decision (which includes
310 the reaction time.)
311
312 Changes in accuracy and latency across the session were measured in response to
313 different stimuli. To assess possible changes and attentional or fatigue effects we obtained
314 the difference between the first 150 and the last 150 measurements accumulated for each
315 subject. Non-parametric statistics were used to assess significance, as the distributions of
316 these parameters were skewed. Furthermore, for the TD children below four years of age
317 and the participants with ASD, the log-transformed of the speed data failed the normality
318 test (Chi-square goodness of fit test in MATLAB rejected the null hypothesis of normality at
319 5% significance level.) This result differed from the TD adults, whose velocity dependent
320 estimated distribution was log-normal.

321 Distributional Analyses


322 Distributional analyses were performed on both sets of parameters. In prior work
323 we had discovered that the two-parameter continuous Gamma family of probability
324 distributions captures with high confidence the continuum of human motor signatures in
325 both typical and compromised systems (Torres 2012; Torres 2013a; b; Torres 2011; Torres
326 et al. 2013a; Yanovich et al. 2013). Thus we used the Gamma plane here to plot the (shape,
327 scale) Gamma parameters for each individual as they were experimentally estimated with
328 95% confidence intervals. The movement decision latency values were also used to assess
329 whether there were marked differences within each sex and between males and females
330 during the match to sample task. To assess the dispersion of the estimated distribution we
331 estimated the mean and variance and obtained the Fano Factor (Fano 1947), the ratio of the

10
σ w2
332 variance divided by the mean, (noise to signal ratio) F = taken within the time window
μw
333 between the movement onset and the peak velocity. For the Gamma probability distribution,
334 μˆ = a ⋅ b and σˆ = a ⋅ b2 , where a is the estimated shape parameter and b is the estimated
335 scale parameter. This ties the scale parameter to the Fano Factor (dispersion.) The lower the
336 value of b, the lower the dispersion (i.e. the more reliable the probabilistic process is.) The
337 higher the value of the shape parameter, the higher the predictability of the process is. Thus,
338 values down and to the right of the Gamma plane are ideal, meaning more reliable and
339 predictable patterns. Likewise, points up and to the left of the Gamma plane are more
340 random and noisier. In the domain of motor signatures taken as re-afferent feedback, values
341 up and to the left of the Gamma plane would not provide the type of motor feedback able to
342 accumulate evidence towards a verifiable expectation that could eventually turn into an
343 anticipatory motor percept.

344 Principal Component Analyses of the percent of time to the maximum speed
345 To assess possible differences in the variability of the timing between forward and
346 withdrawing segments of the reach across subjects we divided the sets into TD and ASD,
347 sorted by age.
348 (1) The data set was organized as an m x n matrix, where m was the number of
349 participants (28 girls, 44 boys) and n the number of samples per participant (50).
350 (2) The estimated mean was subtracted for each measurement type to center the data at
351 zero-mean;
352 (3) The Singular Value Decomposition (SVD) was calculated, obtaining the eigenvectors
353 of the covariance matrix.
354 (4) The principal components were plotted in matrix form for the TD and ASD groups to
355 visualize potential differences in timing variability between the intended and
356 unintended motions.

357 Results
358 All individuals with ASD had highly skewed distributions with a tendency to slower
359 motions and more random and noisier patterns over time (up and to the left of the Gamma
360 plane), as shown in Fig.2B. The verbal females with ASD separated from the non-verbal
361 females in Fig. 2D according to the empirically estimated probability distributions for each
362 individual when using the experimentally determined range of maximum speed values from
363 the data set. Non-verbal and verbal females differed significantly in the dispersion of their
364 distributions according to the ranksum test of the individual’s Fano Factor (p<0.03 forward
11
365 and p<0.01 withdrawing movements.) The dispersion in the distributions of non-verbal
366 females with ASD were close to that of the TD individuals younger than four years old
367 (estimated across thousands of values, p<0.38) shown in Fig.2C. The verbal females with
368 ASD differed from the TD females older than four (p<0.035) and also from the TD females
369 younger than four (p<0.023.) According to the estimated probability distributions, three
370 distinct female groups appeared: TD, verbal ASD and non-verbal ASD, thus highlighting the
371 importance of establishing female normative data to assess atypical traits in females (rather
372 than screening females using the male phenotype as reference, which is the current
373 practice.)
374
375 In the ASD males below five years of age, no clear distinction appeared between
376 these and the TD male children below four years of age in the dispersion of the empirically
377 estimated probability distribution function (estimated across thousands of trials, forward
378 p<0.57, backwards p<0.87); between these and the 4-5 year old TD children (estimated
379 across thousands of trials, forward p<0.26; withdrawing p<0.82). No distinctions were
380 found between the verbal and non-verbal male participants with ASD regardless of age
381 (estimated across thousands of values, forward p<0.61; withdrawing p<0.89) in contrast to
382 the differences found between verbal and non-verbal females with ASD. The kinesthetic
383 input that the speed-dependent variability may provide in the participants with ASD is in
384 general noisy and lacks the diversification quantified in the TD children older than four (Fig.
385 2D-E.)
386 The dispersion comparisons between the 10 females with ASD and 24 males with
387 ASD revealed no significant differences according to the rank sum test (estimated across
388 thousands of measurements, forward p<0.53; withdrawing p<0.65). It would be important
389 to increase the number of subjects in these types of comparisons before drawing further
390 conclusions about inter-group differences.
391
392 Overall, this velocity dependent metric revealed marked differences in the
393 empirically estimated probability distributions between females with ASD and TD females,
394 even at an early age. They also revealed differences between non-verbal and verbal females
395 with ASD that were absent when comparing verbal and non-verbal males with ASD.
396 Furthermore, no differences appeared between the females and males with ASD regardless
397 of age and verbal abilities, but the latter result should be interpreted with caution due to
398 small sub-sample sizes. The most effective comparisons were between TD females and those
399 with ASD, particularly in their (unintended) withdrawing motions, and also between TD

12
400 males and those with ASD. These results strongly suggest that the sexes should be studied
401 separately when assessing movement-based kinesthetic processing in ASD.

402 Normative data: Typical sex differences between forward and withdrawing segments
403 The percent of time to reach the maximum speed differed between forward and
404 withdrawing segments for TD females of college age, but not for the corresponding males.
405 This is shown in Fig.3A-B where the points scatter about the lines, which align differently in
406 females than in males. They have different slopes and intercepts characterizing the two sex
407 groups. The dispersion of the empirically estimated distributions was found to be different
408 as well. Timing was found to be more variable in the females (Fano Factor ratio ranging from
409 0.46 to 7.28), spanning a broader range of values (reaching the speed maximum in the range
410 30% to 56% of the movement duration) and unambiguously differentiating between
411 intended and unintended motion segments (Fig. 3C). Recall here that motions deliberately
412 aimed at the target were intentionally performed but the retractions were uninstructed.
413 These motions are differentially treated by the female sensory-motor systems when
414 compared to the males’, a result that is very relevant in light of the objective nature of this
415 metric. The movements of the females distinguished on average between the forward % of
416 time to peak velocity (0.44 +/- 0.0023) and withdrawing segments (0.34 +/-0.0063; out of
417 1). In marked contrast, the males had less variability (Fano Factor ranging between 0.87 and
418 4.83) and a narrower bandwidth of parameter ranges (reaching the maximum speed at 37-
419 40% of the movement duration) that did not distinguish the timing of the speed maxima in
420 intended and unintended reaches as well as the females did. Forward motions on average
421 reached the maximum speed at (0.46 +/- 0.0012 out of 1) and in the retractions did at 0.40
422 +/- 0.0096. The differences between TD females and males were also evident in the slopes
423 and intercepts of the power-law fit reported in the Fig. 3 caption.
424
425 These differences in the statistical signatures of the variability of the maximum
426 speed timing for TD participants led us to further assess the variability of this kinematic
427 parameter for participants with ASD. To this end, we subtracted the mean from each
428 measurement type to center the data at zero-mean and then calculated the Singular Value
429 Decomposition (SVD) to obtain the eigenvectors of the covariance matrix (Daffertshofer et
430 al. 2004). The results of the principal component analyses (PCA) thus performed are
431 depicted in matrix form in Fig. 4. No apparent differences using PCA were found in the
432 forward reaches intended to touch the target. However, in the uninstructed withdrawing
433 segments marked differences were observed in the ways in which the data self-clustered
434 according to the principal components (PC) across groups. Three clusters appeared for the

13
435 females. From rows 1-6, the younger TD females below five years of age self-aggregated and
436 separated from the older TD females above seven years of age (rows 7-17). In rows 18-28,
437 we see the variability of all the females with ASD (4-25 years of age.) which resembles that
438 of the TD girls below 4 years of age, even though most females with ASD were older than five
439 years of age. This confirmed the previous results concerning the peak velocity (Fig. 2) in the
440 match to sample task between the young TD participants below four years of age and the
441 participants with ASD. The PCA results suggest that the patterns of maximum speed timing
442 in females with ASD also fail to transition into the levels of statistical variability that the TD
443 controls attain after four years of age.
444
445 Parenthetically the PCA yielded an unexpected result. Two exceptions appeared in
446 the TD group that clustered with the groups of participants with ASD. These 2 participants
447 were child prodigies and one of them had received a diagnosis of ASD at 3 years of age but
448 was considered “normal” at present. The spontaneous motions however, blindly situated
449 both subjects within the ASD groups with respect to the latency of the peak velocity. This
450 result came as a surprise, although both subjects scored very high in another study on the
451 Autism Quotient questionnaire by Baron-Cohen (Baron-Cohen et al. 2001). Interestingly,
452 both subjects are at present theoretical physicists (as that resonates with Baron-Cohen’s
453 proposition that people in the hard core, exact sciences tend to have social autistic traits.)
454
455 The lack of distinction between the principal components in the variability of the
456 latency of the maximum speed between TD and ASD participants for the forward reach was
457 intriguing in light of the highly informative patterns provided by the normalized values of
458 the maximum speed. For the females, marked differences in the dispersion of their
459 empirically estimated distributions of the time to the maximum velocity were found
460 between TD and ASD groups across ages (3-4 years old TD vs. low-functioning ASD, p < 0.03
461 ranksum test on the Fano Factor; 4-5 years old TD vs. low-functioning ASD, p< 0.01; both
462 young TD groups also differed from the high-functioning females with ASD p<0.01 and
463 overall the low functioning and the high functioning females with ASD differed from the
464 young TD adults, p < 4.5 x 10-5). These differences were not observed for both forward and
465 withdrawing motions when we compared the females with ASD and the males with ASD in
466 both verbal and non-verbal participants (estimated across thousands of measurements, low-
467 functioning non-verbal, p<0.34; high-functioning verbal, p<0.89). There were no significant
468 differences with respect to the dispersion of the distributions of the TD males younger than
469 4 years old and that of the non-verbal, low-functioning females with ASD (estimated across
470 thousands of values, p<0.17), but there were differences in the dispersion between the TD
14
471 males younger than four years old and that of the verbal, high-functioning females with ASD
472 (p<0.015). A larger number of subjects will be needed to re-assess inter-group comparisons
473 where no significant differences in dispersion were found.
474
475 An important distinction between the latencies of the forward and withdrawing
476 motions should be made: They were on different time scales. Forward motions were slower
477 than the retractions (ranksum test p<10-6). In the forward reach, the median time to the
478 maximum speed was between 172.95ms and 210.53ms for those with ASD and between
479 109.50ms and 179.81ms for TD participants. In contrast, that for the withdrawal was
480 between 93.28ms and 108.30ms for participants with ASD and between 60.34ms and
481 152.11ms for TD.
482
483 These results further motivated us to compare the latency of the movement decision-
484 time (DT) during the match-to-sample task, which may also relate to the kinesthetic sensing
485 of the speed timing, and to the crosstalk between sensory input from vision-touch and
486 sensory input from the unfolding motions.

487 Movement Decision Time (DT)


488 Using the statistical self-emerging clusters obtained in Fig. 2 as a guide and those in
489 (Torres et al. 2013a), the DT frequency histograms were obtained for 6 clusters, 3 TD and 3
490 ASD.
491 We observed highly skewed distributions (Fig. 5) for this decision movement latency
492 parameter, measured in seconds, (5A) with differences between clusters that were well
493 captured by the MLE of the Gamma parameters and plotted on the Gamma plane, with 95%
494 confidence intervals also plotted for each cluster in 5B. In the ASD groups the clusters
495 comprised non-verbal 4-8 year-old children (including males and females) with IQ below
496 50; non-verbal 8-16 year old participants with IQ above 50, and verbal participants 10-25
497 years of age with IQ above 90. The TD clusters were verbal 3-4 year old children with
498 estimated IQ of around 100; verbal children ages 4-5, IQ of around 100, as well as college
499 level fellows 21-30, IQ over 100. These subjects performed both the biomechanical pointing
500 task and the match-to-sample task involving the pointing decisions.
501
502 We also examined the log-transformed data, which had a Gaussian tendency, to assess
503 potential sex dependency and the effects of cognitive load. However, the log-transformed
504 data failed the χ2 goodness of fit test for normality (χ2 = 20.11, p < 10-4), so we used the non-
505 parametric Kruskal-Wallis (non parametric ANOVA) test. The groups were divided by sex to

15
506 evaluate whether the DT distributions differed on this dimension. Differences between
507 groups were found across stimulus conditions for color (χ2 = 57.16, p < 10-13, 1,007samples),
508 shape (χ2 = 117.6, p < 10-26, 1,199 samples), and rotation (χ2 = 168.06, p < 10-37, 1,463
509 samples).
510
511 Further changes in the decision-making parameters reflected the motor adaptation
512 process that accompanied the increases in the cognitive load of the visual stimuli. Percent
513 correct is the total percentage of correct responses for all individuals for each stimulus type.
514 Decision Time (DT) is the length of time from stimulus presentation to when the participant
515 touched one of the two targets. DT reduction is the average number of milliseconds by which
516 participants got faster when comparing their performance for the first 50 trials to the last 50
517 trials of the task for each stimulus type. This gives a measure of performance gains over
518 time. The performance of early vs. late trials was compared to assess changes in decision-
519 making performance as impacted by the changes in motor performance as the speed profiles
520 were adapted and changed with the cognitive stimuli from multi-peaked to unimodal.
521 Number of measurements used across stimulus types and subjects included (2,546 color;
522 3,588 shape; 2,996 rotation).
523
524 The participants with ASD experienced a reduction in the movement decision time (DT)
525 between earlier and later trials, thus ruling out fatigue or boredom. Color was the stimulus
526 type with the fastest decision (reduction of 2,285ms on average (2,370 ms standard
527 deviation); followed by shape 2,570.8 ms (2,926.9ms) and rotation 2,739.8ms (3,527.8ms).
528 Statistical comparison using Kruskall-Wallis revealed significance in ASD 6.7x10-7, χ2 =
529 28.41; 0.003, χ2 = 11.3; 0.05, χ2 = 5.87 for color, shape and rotation respectively. In the TD
530 participants this was also the case 4.5x10-5, χ2 = 20; 1.2x10-9, χ2 = 41.02; 2.1x10-13, χ2 = 58.37,
531 with mean (+/- standard deviation) reductions in ms of 1,556 (1,048); 1,757 (1,480); 2,798
532 (3,325) for color, shapes and rotation respectively.
533
534 For the individual self-emerging clusters, these changes were also significant across
535 conditions for cluster 1 (non-verbal ASD 4-6 years old, p < 0.0002, χ2 = 17.97) and cluster 2
536 (non-verbal ASD 8-16 years old, p < 5.8 x 10-5), but not significant for cluster 3 (verbal ASD
537 10-25 years old, p < 0.80, χ2 = 0.45). The TD cluster 4 (kindergarten) also had a significant
538 reduction in the latency of the decision-making motion (TD 3-4 years old, p<0.0003 χ2 16.53)
539 but not significant in the pre-school cluster 5 (TD 4-6 years old, p < 0.49, χ2 = 1.42) and in the
540 college level cluster 6 (21-30 years old, p < 0.52, χ2 =1.28).
541
16
542 Tukey’s HSD post-hoc tests revealed that each group was different from each other
543 for both percentage correct and decision time. The direction of significance reveals that
544 rotation was the most difficult task (fewer correct, longer decision time), whereas color was
545 the simplest of the three tasks. P-values (percent correct; DT (ms)) were (0.0001; 0.001)
546 color vs. shape; (0.05; 0.0001) shape vs. rotation; (0.0001; 0.0001) color vs. rotation.
547
548 The MLE of the shape and scale Gamma parameters are shown in Fig.6 on the Gamma
549 parameter plane. Panels 6A and 6B compare within each male and female cluster,
550 respectively, performance for TD vs. ASD groups. The insets are from box plots of the
551 Kruskal-Wallis DT(s) comparison taken across thousands of measurements, which was not
552 significantly different for young males in A, but was significantly different for young females
553 shown in B at the 0.01 alpha level. Panels C-D compared males vs. females within each of the
554 ASD and TD classes. Stronger differences were found between younger TD boys and girls,
555 with a trend that continuously changed from exponential-like to skewed, to Gaussian-like
556 distributions in TD. In contrast, females with ASD showed more randomness in DT
557 distributions with more exponential underlying distributions than males, particularly earlier
558 in life.
559 These results underscore once again the need to examine females with ASD in relation to
560 TD females. For all the parameters used in our investigation, the differences were larger
561 between TD and ASD groups within sex than between the sexes for a given group. Marked
562 differences in variability statistics of the movement-based kinesthetic and decision making
563 parameters were more evident for females with ASD in comparison to TD females than
564 when compared to males with ASD. More importantly, these differences are quite
565 pronounced for the youngest females with ASD that were examined.

566 Decision accuracy in the match-to-sample task


567 Overall, participants chose the correct target 91.0% of the time. Comparing accuracy
568 between groups showed that the accuracy for the TD group was 92.4%, 86.9% for the
569 younger ASD group, and 92.0% for the older ASD cohort. A Kruskal-Wallis ANOVA revealed
570 an effect of the group on accuracy (χ2 = 30.72, p < 0.001, 4,976 trials). Post hoc tests revealed
571 that the younger ASD group was less accurate than the older ASD groups (p < .001), and less
572 accurate than the TD (p < .001) group; however, the TD group did not differ from the older
573 ASD group. Females (93.7%) were more accurate than males (89.3%): χ2 = 27.67, p < 0.001,
574 4,976 trials.
575

17
576 The cognitive load of the discrimination task was partially revealed by comparing
577 overall accuracy for the three stimulus conditions: color (95.2%), shape (94.5%), and
578 rotation (83.4%). A Kruskal-Wallis test revealed an effect of condition-type accuracy (χ2 =
579 175.94, p < 0.001, 4,976 trials.) Post hoc tests revealed that accuracy was significantly less
580 for the rotation condition than for both the color (p < .001) and the shape condition (p <
581 .001); however, the color and shape conditions did not differ from each other. The percent
582 correct in the last portions of the sessions remained, here reported as mean (+/- standard
583 deviation): color (97 (+/-18)), shape (95 (+/-21) and rotation (91(+/-29). The variability
584 was at its highest for rotated stimuli and this was also the condition with the lowest
585 accuracy.
586
587 An observation regarding accuracy of the decision making is that the underlying
588 statistics of the kinematics for the motion executing the decision differed dramatically
589 between ASD and TD groups despite overall similarities in accuracy between the two
590 groups. Thus, different mechanisms must underlie their respective accuracies. It is possible
591 that the TD group can make a better use of their predictable proprioception. Yet in light of
592 the noisy, narrow, and random movement-based kinesthetic input of the ASD individuals,
593 we suggest that they may be relying on the physical spatial properties of the visual stimuli
594 more than on their kinesthetic sensing of velocity dependent fluctuations to attain decision
595 accuracy. Further research investigating this proposition is warranted.

596 Discussion
597 In this work, possible sex differences between participants with ASD and TD participants
598 were investigated based on new metrics that assess movement variability over time as
599 manifested in the continuous flow of motions. This information is taken as a form of re-
600 afferent proprioceptive input, partly contributed by the peripheral nervous systems. We also
601 assessed differences based on the variability of decision making parameters that may be
602 centrally controlled. The stochastic signatures of velocity-dependent parameters as
603 impacted by the cognitive load of visual stimuli for decision making were individually
604 investigated across a highly heterogeneous group of participants with a diagnosis of ASD
605 and compared to TD controls. We assessed both deliberate motions towards a target and
606 spontaneously occurring motion segments that were not instructed and that occurred
607 largely beneath the participant’s awareness. In the pointing experiments where decision
608 making was also required, we measured the latency and accuracy of the decisions. Our
609 hypothesis that differences in the stochastic signatures of velocity-dependent kinematics
610 would exist in ASD and be critical to differentiate sex-based manifestations of ASD received
18
611 support from the data. In particular, the data revealed normative values for TD females, to
612 be used as a basis of comparison in determining atypical traits in females. This is in contrast
613 to current inventories which use the male phenotype as the reference for diagnosis and miss
614 traits that are particular to the females, thus leading to the disparate diagnosis ratio of
615 nearly 5 males per 1 female.
616
617 The velocity-dependent data showed a non-trivial statistical transition during typical
618 development occurring after 4 years of age and marked by a decreased noise to signal ratio
619 that was absent in the ASD participants. The ASD subjects never transitioned into a more
620 reliable motor signature, one that became verifiable and diversified with age. The
621 parameters of empirically estimated velocity-dependent probability distributions for the
622 participants with ASD resembled those of the TD children below four years of age. In the
623 case of females with ASD, a marked distinction was found in the dispersion of their
624 empirically estimated probability distributions which separated verbal from non-verbal
625 participants, a pattern that we did not find in males with ASD (despite the larger sample
626 size.)
627
628 The uninstructed motions of the hand, those occurring spontaneously and below
629 intentional awareness were the most informative in the data. In particular, the timing of
630 their maximum speed from segment to segment separated the ASD participants from the TD
631 fellows. The latency of the maximum speeds of the withdrawing reaches was on the order of
632 30-40% of the path within a time window of 60-150ms. These critical landmarks occurred
633 on very short time scales in contrast to those of the forward reaches during the deliberate
634 processing of target information (engaging vision, touch, and pressure) above 200ms. In this
635 regard, we propose that different general somatic afferent pathways (GSA) may be
636 mediating proprioceptive information for the types of “approach-avoidance” motions under
637 study here. Forward motions revealing marked differences in the statistics of intended
638 speed values may be routed through GSA pathways that reach somatosensory cortex via the
639 thalamus (O'Rahilly and Müller 1983). Withdrawing motions incidental to the task and
640 automatically occurring at significantly shorter time scales, largely beneath intentional
641 awareness, may be routed through phylogenetically older pathways to the cerebellum and
642 other older central sub-cortical structures comprising the limbic system and the striatum.
643
644 It is possible that the early developmental glitch that leads to autism in each individual
645 disrupts first the more primitive systems, those which appeared first in evolution. Adaptive
646 compensatory mechanisms may then change the typical development of proper feedback
19
647 loops towards and within the central commanding areas of the neocortex and result in
648 different coping control mechanisms. Such coping strategies would allow the system to
649 survive but would impede the balance between voluntary and automatic control of behavior.
650 This would result in poor body autonomy and self-regulation, which in turn would impede
651 volitional control. We are at present testing these ideas in the context of new therapeutic
652 concepts for ASD.
653
654 Evidence in support of these ideas comes from various disconnected sources of research
655 that need to be integrated. In idiopathic autism, older central core structures of the brain are
656 known to be disrupted. These include the amygdala, striatum and cerebellum (Amaral et al.
657 2008; Courchesne 1997; 1991; Mostofsky et al. 2009; Nayate et al. 2005; Schumann et al.
658 2004). In some forms of autism of known etiology, GSA pathways conducting movement,
659 pressure, pain and temperature information are also known to differ by virtue of
660 dysfunctional proteins that are essential for postsynaptic scaffolding (e.g., SHANK 3 in
661 Phelan-McDermid syndrome (Phelan and Rogers 1993)). Such disruptions give rise to
662 stochastic motor patterns similar to those seen in this report (Torres et al. 2013b). Some
663 manifestations of the phenotype are high tolerance to pain, temperature dysregulation, lack
664 of fear, and lack of a general sense or awareness of body in space and time. These individuals
665 also go on to receive a diagnosis of ASD because such afferent anomalies disrupt social and
666 communicative abilities (Aldinger et al. 2013; Phelan and McDermid 2012; Phelan and
667 Rogers 1993; Phelan 2008; Phelan et al. 2001; Strenge et al. 2008; Uchino and Waga 2013).
668 Reports from parents, self-advocates, and neurologists about individuals with idiopathic
669 autism describe symptoms that concord with the phenomenology of this syndrome of
670 known etiology (Damasio and Maurer 1978; Donnellan and Leary 1995; Maurer and
671 Damasio 1982; Robledo et al. 2012). This suggests that at least in part the noisy and random
672 motor patterns uncovered here may contribute to noisy re-afferent feedback. However,
673 further research will be necessary to definitely establish the origins of such general patterns
674 in ASD, and to separate the motor from the sensory components in the velocity dependent
675 signal.
676 Given the present sex differences in movement parameters and the finding that these
677 differences manifest maximally when females are compared to normative female data, it will
678 be important to systematically examine the physiology of peripheral and sub-cortical areas
679 involved in the feedback loops for proper motor control so as to determine possible
680 physiological sex differences that would explain those unveiled here in the motor readout. In
681 this regard several mouse models have been created yielding phenotypes that involve social
682 impairments similar to autistic-like symptoms (El-Kordi et al. 2012; Giza et al. 2010; Jamain
20
683 et al. 2003; Jamain et al. 2008; Peca et al. 2011; Tabuchi et al. 2007; Wang et al. 2011), yet no
684 objective measurements of their behavior has been provided beyond descriptive
685 observations. Given that motor variability is also present in the animal behaviors; our
686 framework offers a new avenue for objective quantification of sex differences in the
687 physiology of motor control using animal models of ASD.
688
689 It will be important in our future research at the behavioral level to further disentangle
690 and refine the statistical differences in movement variability that we have found between
691 different functional classes of movements, particularly with respect to differing
692 manifestations of these in males and females with ASD. The phylogenetically different GSA
693 pathways may hold promise in shedding light on the much needed characterization of the
694 phenotype of females with ASD because, evolution endowed males and females with
695 different endocrine systems bound to impact differently the somatosensory capacities and
696 their maturation rates. The autistic system, as any biological system, has adaptive
697 capabilities and changes over time. One of the advantages of the present metrics is that they
698 enable the individualized dynamic tracking in real time of the non-stationary statistics of
699 natural movements as a function of age and developmental stage. Thus we could track the
700 same individual longitudinally and chart out the developmental trajectory of the stochastic
701 signatures of his/her motor readout of some aspects of internal proprioception (Torres et al.
702 2013a).
703 In tandem with the kinesthetic-related parameters, decision-making parameters such as
704 the latency and the accuracy of the decision were also examined here. These analyses
705 revealed marked differences between males and females with ASD. Importantly, as with the
706 kinematics data, the differences were more remarkable when examining females with ASD
707 relative to TD females. This trend was more pronounced for younger subjects (even within a
708 small sample size) yet the differences also manifested in the older females with ASD. Our
709 results may be important in light of the present inventories that lack provisions to
710 characterize the female phenotype and rather use the male phenotype as the reference. The
711 most important finding of this study is precisely the revelation that normative data is
712 different between TD males and TD females and that when looked at relative to this
713 normative data, the youngest females with ASD in the group are maximally different.

714 Conclusion and Future Steps


715
716 Although disturbances in movement proprioception are not part of the core symptoms
717 that define ASD, and although most research focuses on centrally-based intentionality,
21
718 planning, and subjective inferences, it may be worthwhile to turn our attention to the body
719 physiology in ASD and to the potential contributions of afferent peripheral inputs from the
720 somatic, autonomic, and enteric nervous systems to the formation of mental operations
721 involved in social exchange.
722
723 The present results on sex-based differences in the statistics of velocity-dependent
724 motor patterns invite further examination of the movement variability of individuals with
725 ASD in relation to TD controls. We suggest that the patterns uncovered here may reveal
726 marked developmental differences in re-afferent kinesthetic sensing between individuals
727 with ASD and controls. To test this proposition in future work, it will be important to
728 separate potential contributions to action planning and control by re-afferent sensory
729 prediction from contributions by the efference copy of ongoing centrally driven motor
730 commands. Both components of internal models for action may be disrupted in ASD, where
731 reports of low muscle tone, high tolerance to pain and temperature dysregulation abound.
732 These anomalies may also contribute to their lack of body awareness, emotional
733 dysregulation and lack of volitional control of actions on command. In this regard, it has
734 been our finding that in ASD, spontaneous motions occurring largely beneath awareness
735 have patterns of variability closer to those quantified in typical controls (Torres 2012;
736 Torres et al. 2013a; Torres et al. 2013d). However, the stochastic signatures of peak velocity
737 and acceleration from intentional segments of the reach in our participants with ASD (both
738 verbal and non-verbal) align well with the patterns of deliberate pointing from a patient
739 without proprioception (Torres et al. 2012), thus suggesting that at least in part, corrupted
740 proprioception may underlie some of the anomalies in volitional control. There have been
741 previous reports of no proprioceptive deficits in high-functioning verbal individuals with
742 ASD (Fuentes et al. 2011; Haswell et al. 2009; Izawa et al. 2012) but studies of non-verbal, so
743 called “low-functioning” individuals have been lacking. In particular those involving analyses
744 of kinematics in verbal and non-verbal females with ASD were nonexistent.
745
746 We have been able to use in children with ASD the spontaneous variability present in
747 natural behaviors to evoke and sustain intentionality in their actions, so as to shift their
748 random and noisy motor patterns into predictable and reliable regimes (Torres et al.
749 2013d). The spontaneous self-discovery of cause-effect by these individuals, in the absence
750 of explicit instructions, has helped them connect their intentions to their actions and retain
751 positive gains over time. These gains and retention are absent in other therapeutic
752 interventions (Black et al. 1972; Cooper et al. 1987) that explicitly command the individual
753 under the hidden assumptions that top-down regulation in ASD exists (from neo-cortex to
22
754 sub-cortical to peripheral pathways) and develops as in the typical participants. TD
755 participants can follow command, have body self-autonomy and self-regulate. In ASD none
756 of these basic functions are typical. We have proposed that new interventions in autism,
757 where the peripheral input is used to evoke central control in a bottom-up fashion (from
758 peripheral to sub-cortical to neo-cortical regions), will be more appropriate than
759 approaches that assume that top-down control is intact in ASD. The autistic system may
760 have developed different adaptive-coping strategies to compensate for the lack of or the
761 corrupted motor feedback (Brincker and Torres 2013; Torres et al. 2013a; Torres et al.
762 2013d). In light of the present results on sex specificity we plan to tailor such therapies to
763 females and males sensory-motor timing features.
764
765 Because of the atypical developmental trajectories of their motion patterns and the lack
766 of age-transition into more predictive, reliable and anticipatory patterns, it may be possible
767 that central (cortical-sub-cortical based control) differs in ASD. Perhaps recruiting the
768 peripheral nervous systems in ASD and guiding it through more primitive sub-cortical
769 mechanisms, particularly those involved in affective and emotional regulation, tempo,
770 rhythm, etc. would lead to better focus, self-regulation and eventually better intentional
771 control. Taking this “back door” route in neurological music therapies (Hardy and LaGasse
772 2013) has helped children find autonomy and self-regulate, as suggested also by metronome
773 based interventions and interventions that involve the acquisition by “low-functioning” non
774 verbal individuals with ASD of independent typing and communication abilities through
775 emotional support (Dowden and Marriner 1995; Kasa-Hendrickson et al. 2009; Orlievsky
776 and Curkier 2013). Timing and rhythms seem to be critical to connect with the individual
777 with ASD (Amos 2013; Barnhill 2013). Here the sex differences may help refine such
778 therapeutic interventions and tailor them to harness the capabilities and predispositions of
779 each individual in the spectrum.
780
781 Differences in timing variability objectively quantified here in decision making processes
782 that occurred in tandem with the hand motions were also uncovered in the uninstructed
783 motions performed largely below intentional awareness. These types of motions will open a
784 window into the potential capabilities of non-verbal individuals with ASD, currently largely
785 under studied. Spontaneous acts do not require verbal instructions or their understanding
786 by the participant. Our metrics can also quantify such motions. Overall the differences in
787 decision making parameters revealed precise sex differences. The new framework and
788 simple experimental paradigms that we introduce here could be useful in the near future to
789 aid researchers and clinicians in the tracking of the continuous flow of natural behaviors
23
790 (e.g. (Torres 2013a; b; Torres et al. 2013d).) They could easily and systematically reproduce
791 our results by testing them within each sex group of the general population, and begin
792 strategizing to find the path towards the objective characterization of the female phenotype
793 with ASD.

794 Acknowledgements
795 We thank Dr. Sandy Harris, Dr. Lara Delmolino, Dr. Kate Fiske, Jennifer Biddick, and
796 Meredith Bamond from the Douglass Developmental Disability Center (DDDC) at Rutgers
797 University, and all the staff members at the DDDC for assistance with participants, data
798 collection, and clinical records. We thank Carrie Fisher and Dawn L. Graham for assistance
799 with participants and their medical records at the Institute for Psychiatric Research and also
800 the Riley Hospital for Children, both part of Indiana University School of Medicine. This work
801 was funded by NSF Cyber-Enabled Discovery and Innovation Type I (Idea) grant # 094158
802 to EBT and DNM “A novel quantitative framework to study lack of social interactions in ASD
803 Spectrum Disorders” and by the New Jersey Governor’s Council for Medical Research and
804 Treatment of Autism grant # 10-403-SCH-E-0 “Perceptual Motor Anticipation in ASD”.
805

24
806 References
807 Adam JJ, Paas FG, Buekers MJ, Wuyts IJ, Spijkers WA, and Wallmeyer P. Gender
808 differences in choice reaction time: evidence for differential strategies. Ergonomics 42: 327-335,
809 1999.
810 Aldinger KA, Kogan J, Kimonis V, Fernandez B, Horn D, Klopocki E, Chung B, Toutain A,
811 Weksberg R, Millen KJ, Barkovich AJ, and Dobyns WB. Cerebellar and posterior fossa
812 malformations in patients with autism-associated chromosome 22q13 terminal deletion. Am J Med
813 Genet A 161: 131-136, 2013.
814 Amaral DG, and Corbett BA. The amygdala, autism and anxiety. Novartis Found Symp 251:
815 177-187; discussion 187-197, 281-197, 2003.
816 Amaral DG, Schumann CM, and Nordahl CW. Neuroanatomy of autism. Trends Neurosci 31:
817 137-145, 2008.
818 American Psychiatric Association. Diagnostic and statistical manual of mental disorders :
819 DSM-IV-TR. Washington, DC: American Psychiatric Association, 2000, p. xxxvii, 943 p.
820 Amos P. Rhythm and timing in autism: learning to dance. Front Integr Neurosci 7: 27, 2013.
821 Andersen RA, and Buneo CA. Intentional maps in posterior parietal cortex. Annu Rev Neurosci
822 25: 189-220, 2002.
823 Ashwood P, Anthony A, Pellicer AA, Torrente F, Walker-Smith JA, and Wakefield AJ.
824 Intestinal lymphocyte populations in children with regressive autism: evidence for extensive
825 mucosal immunopathology. J Clin Immunol 23: 504-517, 2003.
826 Bandstra NF, Johnson SA, Filliter JH, and Chambers CT. Self-reported and parent-reported
827 pain for common painful events in high-functioning children and adolescents with autism
828 spectrum disorder. Clin J Pain 28: 715-721, 2012.
829 Barnhill E. Neural connectivity, music, and movement: a response to Pat Amos. Front Integr
830 Neurosci 7: 29, 2013.
831 Baron-Cohen S. The extreme male brain theory of autism. Trends Cogn Sci 6: 248-254, 2002.
832 Baron-Cohen S, Richler J, Bisarya D, Gurunathan N, and Wheelwright S. The systemizing
833 quotient: an investigation of adults with Asperger syndrome or high-functioning autism, and
834 normal sex differences. Philos Trans R Soc Lond B Biol Sci 358: 361-374, 2003.
835 Baron-Cohen S, Wheelwright S, Skinner R, Martin J, and Clubley E. The autism-spectrum
836 quotient (AQ): evidence from Asperger syndrome/high-functioning autism, males and females,
837 scientists and mathematicians. J Autism Dev Disord 31: 5-17, 2001.
838 Berument SK, Rutter M, Lord C, Pickles A, and Bailey A. Autism screening questionnaire:
839 diagnostic validity. Br J Psychiatry 175: 444-451, 1999.
840 Black AH, Gilbert RM, and Millenson JR. Reinforcement; behavioral analyses. New York,:
841 Academic Press, 1972, p. xiii, 288 p.
842 Bourgeron T. The possible interplay of synaptic and clock genes in autism spectrum disorders.
843 Cold Spring Harb Symp Quant Biol 72: 645-654, 2007.
844 Breece E, Paciotti B, Nordahl CW, Ozonoff S, Van de Water JA, Rogers SJ, Amaral D, and
845 Ashwood P. Myeloid dendritic cells frequencies are increased in children with autism spectrum
846 disorder and associated with amygdala volume and repetitive behaviors. Brain Behav Immun
847 2012.
848 Brincker M, and Torres EB. Noise from the periphery in Autism. Frontiers in Integrative
849 Neuroscience 7: 1-4, 2013.
850 Buie T, Fuchs GJ, 3rd, Furuta GT, Kooros K, Levy J, Lewis JD, Wershil BK, and Winter H.
851 Recommendations for evaluation and treatment of common gastrointestinal problems in children
852 with ASDs. Pediatrics 125 Suppl 1: S19-29, 2010.
853 Campbell JM. Diagnostic assessment of Asperger's disorder: a review of five third-party rating
854 scales. J Autism Dev Disord 35: 25-35, 2005.
855 Clark FJ, Matthews PB, and Muir RB. Effect of the amplitude of muscle vibration on the
856 subjectively experienced illusion of movement [proceedings]. J Physiol 296: 14P-15P, 1979.

25
857 Cooper JO, Heron TE, and Heward WL. Applied behavior analysis. Columbus: Merrill Pub.
858 Co., 1987, p. xvii, 651 p.
859 Courchesne E. Brainstem, cerebellar and limbic neuroanatomical abnormalities in autism. Curr
860 Opin Neurobiol 7: 269-278, 1997.
861 Courchesne E. Neuroanatomic imaging in autism. Pediatrics 87: 781-790, 1991.
862 Daffertshofer A, Lamoth CJ, Meijer OG, and Beek PJ. PCA in studying coordination and
863 variability: a tutorial. Clin Biomech (Bristol, Avon) 19: 415-428, 2004.
864 Damasio AR, and Maurer RG. A neurological model for childhood autism. Arch Neurol 35:
865 777-786, 1978.
866 de Magistris L, Familiari V, Pascotto A, Sapone A, Frolli A, Iardino P, Carteni M, De Rosa
867 M, Francavilla R, Riegler G, Militerni R, and Bravaccio C. Alterations of the intestinal barrier
868 in patients with autism spectrum disorders and in their first-degree relatives. J Pediatr
869 Gastroenterol Nutr 51: 418-424, 2010.
870 Der G, and Deary IJ. Age and sex differences in reaction time in adulthood: results from the
871 United Kingdom Health and Lifestyle Survey. Psychol Aging 21: 62-73, 2006.
872 Derbyshire SW, Nichols TE, Firestone L, Townsend DW, and Jones AK. Gender differences
873 in patterns of cerebral activation during equal experience of painful laser stimulation. J Pain 3:
874 401-411, 2002.
875 Donnellan AM, Hill DA, and Leary MR. Rethinking autism: implications of sensory and
876 movement differences for understanding and support. Front Integr Neurosci 6: 124, 2012.
877 Donnellan AM, and Leary MR. Autism : sensory-movement differences and diversity.
878 Cambridge, WI: Cambridge Book Review Press, 2012.
879 Donnellan AM, and Leary MR. Movement differences and diversity in autism/mental
880 retardation : appreciating and accommodating people with communication and behavior
881 challenges. Madison, WI: DRI Press, 1995, p. 107 p.
882 Dowden PA, and Marriner NA. Augmentative and alternative communication: treatment
883 principles and strategies. Semin Speech Lang 16: 140-157; quiz 157-148, 1995.
884 Dubois A, Rattaz C, Pry R, and Baghdadli A. [Autism and pain - a literature review]. Pain Res
885 Manag 15: 245-253, 2010.
886 El-Kordi A, Winkler D, Hammerschmidt K, Kastner A, Krueger D, Ronnenberg A, Ritter
887 C, Jatho J, Radyushkin K, Bourgeron T, Fischer J, Brose N, and Ehrenreich H.
888 Development of an autism severity score for mice using Nlgn4 null mutants as a construct-valid
889 model of heritable monogenic autism. Behav Brain Res 2012.
890 Fano U. Ionization Yield of Radiations. II. The Fluctuations of the Number of Ions. Physical
891 Review 72: 26, 1947.
892 Fournier KA, Hass CJ, Naik SK, Lodha N, and Cauraugh JH. Motor coordination in autism
893 spectrum disorders: a synthesis and meta-analysis. J Autism Dev Disord 40: 1227-1240, 2010a.
894 Fournier KA, Kimberg CI, Radonovich KJ, Tillman MD, Chow JW, Lewis MH, Bodfish
895 JW, and Hass CJ. Decreased static and dynamic postural control in children with autism
896 spectrum disorders. Gait Posture 32: 6-9, 2010b.
897 Freeman JB, and Ambady N. MouseTracker: software for studying real-time mental processing
898 using a computer mouse-tracking method. Behav Res Methods 42: 226-241, 2010.
899 Fuentes CT, Mostofsky SH, and Bastian AJ. No proprioceptive deficits in autism despite
900 movement-related sensory and execution impairments. J Autism Dev Disord 41: 1352-1361, 2011.
901 Georgopoulos AP, Schwartz AB, and Kettner RE. Neuronal population coding of movement
902 direction. Science 233: 1416-1419, 1986.
903 Giarelli E, Wiggins LD, Rice CE, Levy SE, Kirby RS, Pinto-Martin J, and Mandell D. Sex
904 differences in the evaluation and diagnosis of autism spectrum disorders among children. Disabil
905 Health J 3: 107-116, 2010.
906 Gidley Larson JC, Bastian AJ, Donchin O, Shadmehr R, and Mostofsky SH. Acquisition of
907 internal models of motor tasks in children with autism. Brain 131: 2894-2903, 2008.
908 Gilliam J. GARS-2: Gilliam Autism Rating Scale-Second Edition. Austin, TX: PRO-ED, 2006.
909 Giza J, Urbanski MJ, Prestori F, Bandyopadhyay B, Yam A, Friedrich V, Kelley K,
910 D'Angelo E, and Goldfarb M. Behavioral and cerebellar transmission deficits in mice lacking
911 the autism-linked gene islet brain-2. J Neurosci 30: 14805-14816, 2010.
26
912 Glickman G. Circadian rhythms and sleep in children with autism. Neurosci Biobehav Rev 34:
913 755-768, 2010.
914 Gotham K, Pickles A, and Lord C. Standardizing ADOS scores for a measure of severity in
915 autism spectrum disorders. J Autism Dev Disord 39: 693-705, 2009.
916 Gowen E, Stanley J, and Miall RC. Movement interference in autism-spectrum disorder.
917 Neuropsychologia 46: 1060-1068, 2008.
918 Hardy MW, and LaGasse B. Rhythm, movement, and autism: using rhythmic rehabilitation
919 research as a model for autism. Frontiers in Integrative Neuroscience 7: 1-9, 2013.
920 Hashmi JA, and Davis KD. Women experience greater heat pain adaptation and habituation than
921 men. Pain 145: 350-357, 2009.
922 Haswell CC, Izawa J, Dowell LR, Mostofsky SH, and Shadmehr R. Representation of internal
923 models of action in the autistic brain. Nat Neurosci 12: 970-972, 2009.
924 Hauschild M, Mulliken GH, Fineman I, Loeb GE, and Andersen RA. Cognitive signals for
925 brain-machine interfaces in posterior parietal cortex include continuous 3D trajectory commands.
926 Proc Natl Acad Sci U S A 109: 17075-17080, 2012.
927 Hwang EJ, Bailey PM, and Andersen RA. Volitional control of neural activity relies on the
928 natural motor repertoire. Curr Biol 23: 353-361, 2013.
929 Izawa J, Pekny SE, Marko MK, Haswell CC, Shadmehr R, and Mostofsky SH. Motor
930 learning relies on integrated sensory inputs in ADHD, but over-selectively on proprioception in
931 autism spectrum conditions. Autism Res 5: 124-136, 2012.
932 Jacobson R, Le Couteur A, Howlin P, and Rutter M. Selective subcortical abnormalities in
933 autism. Psychol Med 18: 39-48, 1988.
934 Jamain S, Quach H, Betancur C, Rastam M, Colineaux C, Gillberg IC, Soderstrom H, Giros
935 B, Leboyer M, Gillberg C, and Bourgeron T. Mutations of the X-linked genes encoding
936 neuroligins NLGN3 and NLGN4 are associated with autism. Nat Genet 34: 27-29, 2003.
937 Jamain S, Radyushkin K, Hammerschmidt K, Granon S, Boretius S, Varoqueaux F,
938 Ramanantsoa N, Gallego J, Ronnenberg A, Winter D, Frahm J, Fischer J, Bourgeron T,
939 Ehrenreich H, and Brose N. Reduced social interaction and ultrasonic communication in a
940 mouse model of monogenic heritable autism. Proc Natl Acad Sci U S A 105: 1710-1715, 2008.
941 Jansiewicz EM, Goldberg MC, Newschaffer CJ, Denckla MB, Landa R, and Mostofsky SH.
942 Motor signs distinguish children with high functioning autism and Asperger's syndrome from
943 controls. J Autism Dev Disord 36: 613-621, 2006.
944 Jones V, and Prior M. Motor imitation abilities and neurological signs in autistic children. J
945 Autism Dev Disord 15: 37-46, 1985.
946 Kasa-Hendrickson C, Broderick AA, and Hanson D. Sorting out Speech: Understanding
947 multiple methods of communication for persons with autism and other developmental disabilities.
948 The Journal of Developmental Processes 4: 116-133, 2009.
949 Kaushanskaya M, Marian V, and Yoo J. Gender differences in adult word learning. Acta
950 Psychol (Amst) 137: 24-35, 2011.
951 Kawato M, and Wolpert D. Internal models for motor control. Novartis Found Symp 218: 291-
952 304; discussion 304-297, 1998.
953 Klintwall L, Holm A, Eriksson M, Carlsson LH, Olsson MB, Hedvall A, Gillberg C, and
954 Fernell E. Sensory abnormalities in autism. A brief report. Res Dev Disabil 32: 795-800, 2011.
955 Kushak RI, Lauwers GY, Winter HS, and Buie TM. Intestinal disaccharidase activity in
956 patients with autism: effect of age, gender, and intestinal inflammation. Autism 15: 285-294, 2011.
957 Lai MC, Lombardo MV, Pasco G, Ruigrok AN, Wheelwright SJ, Sadek SA, Chakrabarti B,
958 and Baron-Cohen S. A behavioral comparison of male and female adults with high functioning
959 autism spectrum conditions. PLoS One 6: e20835, 2011.
960 Lecavalier L. An evaluation of the Gilliam Autism Rating Scale. J Autism Dev Disord 35: 795-
961 805, 2005.
962 Lleonart J, Salat J, and Torres GJ. Removing allometric effects of body size in morphological
963 analysis. J Theor Biol 205: 85-93, 2000.
964 Lord C, and Bishop SL. Autism Spectrum Disorders: Diagnosis, prevalence and services for
965 children and families. Society for Reseach in Child Development 24: 1-26, 2010.

27
966 Lord C, Risi S, Lambrecht L, Cook EH, Jr., Leventhal BL, DiLavore PC, Pickles A, and
967 Rutter M. The autism diagnostic observation schedule-generic: a standard measure of social and
968 communication deficits associated with the spectrum of autism. J Autism Dev Disord 30: 205-223,
969 2000.
970 MacFabe DF, Cain NE, Boon F, Ossenkopp KP, and Cain DP. Effects of the enteric bacterial
971 metabolic product propionic acid on object-directed behavior, social behavior, cognition, and
972 neuroinflammation in adolescent rats: Relevance to autism spectrum disorder. Behav Brain Res
973 217: 47-54, 2011.
974 Mandy W, Chilvers R, Chowdhury U, Salter G, Seigal A, and Skuse D. Sex Differences in
975 Autism Spectrum Disorder: Evidence from a Large Sample of Children and Adolescents. J Autism
976 Dev Disord 2011.
977 Mannion AF, Dumas GA, Cooper RG, Espinosa FJ, Faris MW, and Stevenson JM. Muscle
978 fibre size and type distribution in thoracic and lumbar regions of erector spinae in healthy subjects
979 without low back pain: normal values and sex differences. J Anat 190: 505-513, 1997.
980 Matthews PB. Where does Sherrington's "muscular sense" originate? Muscles, joints, corollary
981 discharges? Annu Rev Neurosci 5: 189-218, 1982.
982 Matthews PBC editor. Muscle spindles: Their messages and their fusimotor supply. Bethersda,
983 MD: American Physiological Society, 1981, p. 189-228.
984 Maurer RG, and Damasio AR. Childhood autism from the point of view of behavioral
985 neurology. J Autism Dev Disord 12: 195-205, 1982.
986 Maurer RG, and Damasio AR. Vestibular dysfunction in autistic children. Dev Med Child
987 Neurol 21: 656-659, 1979.
988 Mazurek MO, Vasa RA, Kalb LG, Kanne SM, Rosenberg D, Keefer A, Murray DS,
989 Freedman B, and Lowery LA. Anxiety, sensory over-responsivity, and gastrointestinal problems
990 in children with autism spectrum disorders. J Abnorm Child Psychol 41: 165-176, 2013.
991 Minderaa RB, Volkmar FR, Hansen CR, Harcherik DF, Akkerhuis GW, and Cohen DJ.
992 Snout and visual rooting reflexes in infantile autism. J Autism Dev Disord 15: 409-416, 1985.
993 Minshew NJ, Sung K, Jones BL, and Furman JM. Underdevelopment of the postural control
994 system in autism. Neurology 63: 2056-2061, 2004.
995 Molloy CA, and Manning-Courtney P. Prevalence of chronic gastrointestinal symptoms in
996 children with autism and autistic spectrum disorders. Autism 7: 165-171, 2003.
997 Mosimann JE. Size Allometry: Size and Shape Variables with Characterizations of the
998 Lognormal and Generalized Gamma Distributions. Journal of the American Statistical
999 Association 65: 930-945, 1970.
1000 Mostofsky SH, Dubey P, Jerath VK, Jansiewicz EM, Goldberg MC, and Denckla MB.
1001 Developmental dyspraxia is not limited to imitation in children with autism spectrum disorders. J
1002 Int Neuropsychol Soc 12: 314-326, 2006.
1003 Mostofsky SH, Powell SK, Simmonds DJ, Goldberg MC, Caffo B, and Pekar JJ. Decreased
1004 connectivity and cerebellar activity in autism during motor task performance. Brain 132: 2413-
1005 2425, 2009.
1006 Mulliken GH, Musallam S, and Andersen RA. Forward estimation of movement state in
1007 posterior parietal cortex. Proc Natl Acad Sci U S A 105: 8170-8177, 2008.
1008 Nader R, Oberlander TF, Chambers CT, and Craig KD. Expression of pain in children with
1009 autism. Clin J Pain 20: 88-97, 2004.
1010 Nayate A, Bradshaw JL, and Rinehart NJ. Autism and Asperger's disorder: are they movement
1011 disorders involving the cerebellum and/or basal ganglia? Brain Res Bull 67: 327-334, 2005.
1012 Newschaffer CJ, Croen LA, Daniels J, Giarelli E, Grether JK, Levy SE, Mandell DS, Miller
1013 LA, Pinto-Martin J, Reaven J, Reynolds AM, Rice CE, Schendel D, and Windham GC. The
1014 epidemiology of autism spectrum disorders. Annu Rev Public Health 28: 235-258, 2007.
1015 Nordahl CW, Scholz R, Yang X, Buonocore MH, Simon T, Rogers S, and Amaral DG.
1016 Increased rate of amygdala growth in children aged 2 to 4 years with autism spectrum disorders: a
1017 longitudinal study. Arch Gen Psychiatry 69: 53-61, 2012.
1018 Noterdaeme M, Mildenberger K, Minow F, and Amorosa H. Evaluation of neuromotor deficits
1019 in children with autism and children with a specific speech and language disorder. Eur Child
1020 Adolesc Psychiatry 11: 219-225, 2002.
28
1021 O'Rahilly R, and Müller F. Basic human anatomy : a regional study of human structure.
1022 Philadelphia: Saunders, 1983, p. xi, 566 p.
1023 Orlievsky D, and Curkier S. Language, writing and activity disorder in the autistic spectrum.
1024 Frontiers in Integrative Neuroscience 7: 1-3, 2013.
1025 Peca J, Feliciano C, Ting JT, Wang W, Wells MF, Venkatraman TN, Lascola CD, Fu Z, and
1026 Feng G. Shank3 mutant mice display autistic-like behaviours and striatal dysfunction. Nature
1027 472: 437-442, 2011.
1028 Phelan K, and McDermid HE. The 22q13.3 Deletion Syndrome (Phelan-McDermid Syndrome).
1029 Mol Syndromol 2: 186-201, 2012.
1030 Phelan K, and Rogers C. Phelan-McDermid Syndrome. 1993.
1031 Phelan MC. Deletion 22q13.3 syndrome. Orphanet J Rare Dis 3: 14, 2008.
1032 Phelan MC, Rogers RC, Saul RA, Stapleton GA, Sweet K, McDermid H, Shaw SR, Claytor
1033 J, Willis J, and Kelly DP. 22q13 deletion syndrome. Am J Med Genet 101: 91-99, 2001.
1034 Prevosto V, Graf W, and Ugolini G. Cerebellar inputs to intraparietal cortex areas LIP and MIP:
1035 functional frameworks for adaptive control of eye movements, reaching, and arm/eye/head
1036 movement coordination. Cereb Cortex 20: 214-228, 2010.
1037 Prevosto V, Graf W, and Ugolini G. Posterior parietal cortex areas MIP and LIPv receive eye
1038 position and velocity inputs via ascending preposito-thalamo-cortical pathways. Eur J Neurosci
1039 30: 1151-1161, 2009.
1040 Prevosto V, Graf W, and Ugolini G. Proprioceptive pathways to posterior parietal areas MIP
1041 and LIPv from the dorsal column nuclei and the postcentral somatosensory cortex. Eur J Neurosci
1042 33: 444-460, 2011.
1043 Qiu A, Adler M, Crocetti D, Miller MI, and Mostofsky SH. Basal ganglia shapes predict
1044 social, communication, and motor dysfunctions in boys with autism spectrum disorder. J Am Acad
1045 Child Adolesc Psychiatry 49: 539-551, 551 e531-534, 2010.
1046 Racine M, Tousignant-Laflamme Y, Kloda LA, Dion D, Dupuis G, and Choiniere M. A
1047 systematic literature review of 10 years of research on sex/gender and experimental pain
1048 perception - part 1: are there really differences between women and men? Pain 153: 602-618,
1049 2012.
1050 Redgrave P, and Coizet V. Brainstem interactions with the basal ganglia. Parkinsonism Relat
1051 Disord 13 Suppl 3: S301-305, 2007.
1052 Redgrave P, Rodriguez M, Smith Y, Rodriguez-Oroz MC, Lehericy S, Bergman H, Agid Y,
1053 DeLong MR, and Obeso JA. Goal-directed and habitual control in the basal ganglia:
1054 implications for Parkinson's disease. Nat Rev Neurosci 11: 760-772, 2010.
1055 Redish AD, and Touretzky DS. The reaching task: evidence for vector arithmetic in the motor
1056 system? Biol Cybern 71: 307-317, 1994.
1057 Reed P. The Return of the Reflex: Considerations of the Contribution of the Early Behaviorism to
1058 Understanding, Diagnosing, and Preventing Autism. In: New Autism Research Developments,
1059 edited by Mesmere BS. Hauppauge NY Nova Science Publishers, 2007, p. 19-24.
1060 Rinehart NJ, Bradshaw JL, Brereton AV, and Tonge BJ. Movement preparation in high-
1061 functioning autism and Asperger disorder: a serial choice reaction time task involving motor
1062 reprogramming. J Autism Dev Disord 31: 79-88, 2001.
1063 Rinehart NJ, Bradshaw JL, Tonge BJ, Brereton AV, and Bellgrove MA. A neurobehavioral
1064 examination of individuals with high-functioning autism and Asperger's disorder using a fronto-
1065 striatal model of dysfunction. Behav Cogn Neurosci Rev 1: 164-177, 2002.
1066 Robledo J, Donnellan AM, and Strandt-Conroy K. An exploration of sensory and movement
1067 differences from the perspective of individuals with autism. Front Integr Neurosci 6: 107, 2012.
1068 Rogers SJ, Bennetto L, McEvoy R, and Pennington BF. Imitation and pantomime in high-
1069 functioning adolescents with autism spectrum disorders. Child Dev 67: 2060-2073, 1996.
1070 Roid GH. Stanford-Binet Intelligence Scales. Itasca, Ill.: Riverside Pub., 2003, p. 5 pieces.
1071 Roll JP, and Vedel JP. Kinaesthetic role of muscle afferents in man, studied by tendon vibration
1072 and microneurography. Exp Brain Res 47: 177-190, 1982.
1073 Savarese RJ. Moving the field: the sensorimotor perspective on autism (Commentary on
1074 "Rethinking autism: implications of sensory and motor differences," an article by Anne
1075 Donnellan, David Hill, and Martha Leary). Front Integr Neurosci 7: 6, 2013.
29
1076 Schumann CM, Hamstra J, Goodlin-Jones BL, Lotspeich LJ, Kwon H, Buonocore MH,
1077 Lammers CR, Reiss AL, and Amaral DG. The amygdala is enlarged in children but not
1078 adolescents with autism; the hippocampus is enlarged at all ages. J Neurosci 24: 6392-6401, 2004.
1079 Schwartz AB. Direct cortical representation of drawing. Science 265: 540-542, 1994.
1080 Shadmehr R, and Wise SP. The computational neurobiology of reaching and pointing : a
1081 foundation for motor learning. Cambridge, Mass.: MIT Press, 2005, p. xvii, 575 p.
1082 Smith Y, Surmeier DJ, Redgrave P, and Kimura M. Thalamic contributions to Basal Ganglia-
1083 related behavioral switching and reinforcement. J Neurosci 31: 16102-16106, 2011.
1084 Strenge S, Froster UG, Kujat A, Bernhard M, and Merkenschlager A. [Muscular hypotonia,
1085 developmental retardation, speech delay and mildly dysmorphic features: 22q13 deletion
1086 syndrome (Phelan-McDermid Syndrome) as an important differential diagnosis]. Klin Padiatr
1087 220: 318-320, 2008.
1088 Stromswold K. Genetics of spoken language disorders. Hum Biol 70: 297-324, 1998.
1089 Tabuchi K, Blundell J, Etherton MR, Hammer RE, Liu X, Powell CM, and Sudhof TC. A
1090 neuroligin-3 mutation implicated in autism increases inhibitory synaptic transmission in mice.
1091 Science 318: 71-76, 2007.
1092 Takarae Y, Minshew NJ, Luna B, and Sweeney JA. Atypical involvement of frontostriatal
1093 systems during sensorimotor control in autism. Psychiatry Res 156: 117-127, 2007.
1094 Teitelbaum P, Teitelbaum O, Nye J, Fryman J, and Maurer RG. Movement analysis in
1095 infancy may be useful for early diagnosis of autism. Proc Natl Acad Sci U S A 95: 13982-13987,
1096 1998.
1097 Thompson J, and Khalil RA. Gender differences in the regulation of vascular tone. Clin Exp
1098 Pharmacol Physiol 30: 1-15, 2003.
1099 Tordjman S, Anderson GM, Botbol M, Brailly-Tabard S, Perez-Diaz F, Graignic R, Carlier
1100 M, Schmit G, Rolland AC, Bonnot O, Trabado S, Roubertoux P, and Bronsard G. Pain
1101 reactivity and plasma beta-endorphin in children and adolescents with autistic disorder. PLoS
1102 ONE 4: e5289, 2009.
1103 Torres E, Cole J, and Poizner H. Spontaneity in the task-incidental segments of the pointing
1104 gesture emerges when visual guidance or movement imagination replace proprioception. In: The
1105 Society for Neuroscience Annual Meeting. New Orleans, LA: 2012.
1106 Torres EB. Atypical signatures of motor variability found in an individual with ASD. Neurocase:
1107 The neural basis of cognition 1: 1-16, 2012.
1108 Torres EB. The rates of change of the stochastic trajectories of acceleration variability are a good
1109 predictor of normal aging and of the stage of Parkinson’s disease. Frontiers in Integrative
1110 Neuroscience in press: 2013a.
1111 Torres EB. Signatures of movement variability anticipate hand speed according to levels of
1112 intent. Behavioral and Brain Functions 9: 10, 2013b.
1113 Torres EB. Two classes of movements in motor control. Exp Brain Res 215: 269-283, 2011.
1114 Torres EB, Brincker M, Isenhower RW, Yanovich P, Stigler KA, Nurnberger JI, Metaxas
1115 DN, and Jose JV. Autism: The Micro-movement Perspective. Frontiers in Integrative
1116 Neuroscience in press: 2013a.
1117 Torres EB, Heilman KM, and Poizner H. Impaired endogenously evoked automated reaching in
1118 Parkinson's disease. J Neurosci 31: 17848-17863, 2011.
1119 Torres EB, Nguyen J, Suresh C, Yanovich P, and Kolevzon A. Noise from the periphery in
1120 autism spectrum disorders of idiopathic origins and of known etiology. In: The Society for
1121 Neuroscience. San Diego, CA: 2013b.
1122 Torres EB, Quian Quiroga R, Cui H, and Buneo CA. Neural correlates of learning and
1123 trajectory planning in the posterior parietal cortex. Front Integr Neurosci 7: 39, 2013c.
1124 Torres EB, Yanovich P, and Metaxas DN. Give spontaneity and self-discovery a chance in
1125 ASD: Spontaneous peripheral limb variability as a proxy to evoke centrally driven intentional
1126 acts. Frontiers in Integrative Neuroscience 2013d.
1127 Uchino S, and Waga C. SHANK3 as an autism spectrum disorder-associated gene. Brain Dev
1128 35: 106-110, 2013.
1129 Vedel JP, and Roll JP. Response to pressure and vibration of slowly adapting cutaneous
1130 mechanoreceptors in the human foot. Neurosci Lett 34: 289-294, 1982.
30
1131 Volkmar FR, Szatmari P, and Sparrow SS. Sex differences in pervasive developmental
1132 disorders. J Autism Dev Disord 23: 579-591, 1993.
1133 Wang X, McCoy PA, Rodriguiz RM, Pan Y, Je HS, Roberts AC, Kim CJ, Berrios J, Colvin
1134 JS, Bousquet-Moore D, Lorenzo I, Wu G, Weinberg RJ, Ehlers MD, Philpot BD, Beaudet
1135 AL, Wetsel WC, and Jiang YH. Synaptic dysfunction and abnormal behaviors in mice lacking
1136 major isoforms of Shank3. Hum Mol Genet 20: 3093-3108, 2011.
1137 Williams JH, Whiten A, Suddendorf T, and Perrett DI. Imitation, mirror neurons and autism.
1138 Neurosci Biobehav Rev 25: 287-295, 2001.
1139 Wolpert DM, Miall RC, and Kawato M. Internal models in the cerebellum. Trends Cogn Sci 2:
1140 338-347, 1998.
1141 Yanovich P, Isenhower RW, Sage J, and Torres EB. Spatial-orientation priming impedes rather
1142 than facilitates the spontaneous control of hand-retraction speeds in patients with Parkinson's
1143 disease. PLoS ONE in press: 2013.
1144 Zeidan-Chulia F, Gursoy UK, Kononen E, and Gottfried C. A dental look at the autistic
1145 patient through orofacial pain. Acta Odontol Scand 69: 193-200, 2011.
1146
1147
1148

31
1149 Figure Captions
1150

1151

1152 Figure 1
1153 Navigating through the continuous flow of natural motions and separating goal-
1154 directed from goal-less segments of behavior.
1155 (A) Hand movement trajectories from a typical child collected during 16.66 seconds of the
1156 match-to-sample task (240 frames/s) which required deciding between two stimulus
1157 choices. The blue circles mark the speed minima (pauses) while the black stars mark the
1158 speed maxima. The black curves denote the pointing trajectories to the green target at two
1159 different positions on the monitor facing the child. The blue curves are incidental to the task,
1160 goal-less movements in transition to other goal-directed motions. (B) Corresponding speed
1161 profiles along the trajectories in A. Numbers and colors correspond to the curves in A. The
1162 speed temporal profile permits to navigate through the acceleration and deceleration phases
1163 of the continuous flow of motion. (C) Zooming into the goal-directed speeds and (D) the

32
1164 goal-less speed profiles which were automatically harnessed by a computer interface (see
1165 methods in the main text).
1166

1167

1168 Figure 2
1169 Typical and atypical development of the statistical patterns of velocity-dependent
1170 variability. The stochastic signatures of velocity-dependent variability captured in the
1171 normalized maximum speed (Vmax / (Vmax + Vavrg)) of pointing motions across different ages.
1172 The estimated shape and scale parameters of the continuous Gamma probability
1173 distribution family uniquely label each individual in the group (78 participants total). (A) 44
1174 typical controls automatically cluster by age along the line of unity on the log-log Gamma
1175 plane according to a scaling power law (circles represent forward segments and diamonds
1176 represent withdrawing segments). (B) 34 participants with ASD also align on the line of
1177 unity. Notice that the 34 participants with ASD include verbal and non-verbal subjects. They
1178 span from 4 to 25 years of age, yet they all fall along the statistical region of the TD 3-4 years
1179 old. (C) Estimated probability distributions of the velocity-dependent parameter for all
1180 typical subjects using empirically obtained hand pointing speed ranges. Notice that the noise
1181 to signal ratio (variance/mean) changes dramatically from 3-4 to 4-5 years of age, along
1182 with the bandwidth of parameter values that the distribution spans across subjects
1183 (significant differences p < 10-5). The 3-4 year old children all collapsed on the same curve
1184 (non-diversified mean) with the noise overpowering the signal, but the children from 4-5

33
1185 years old have acquired a verifiable kinesthetic percept with significantly lower noise to
1186 signal ratio and a broader bandwidth of parameter values (details in the main text). Such
1187 diversification of the kinesthetic input is maximal in the adults who have highly reliable and
1188 predictive statistics of velocity-dependent variability, a form of kinesthetic input. (D-E) The
1189 participants with ASD manifest statistical features of the 3-4 year old TD, unreliable and
1190 noisy with narrow bandwidth of parameter values. The verbal females separate from the
1191 non-verbal females with ASD, but do not quite reach the level of kinesthetic input reliability,
1192 prediction and diversification of the 4-5 year old TD. Unlike the females, the statistical
1193 signatures of the velocity-dependent variability in the males with ASD cannot distinguish
1194 between verbal and non-verbal participants.

1195

1196 Figure 3
1197 Normative typical sex differences for typical young adult in the percent of time to
1198 reach the speed maximum. (A) The patterns of variability of the percent time to reach the
1199 maximum speed in females of college age can distinguish between forward and withdrawing
1200 segments of the reaching movements. Female forward segments can be well fit with
1201 f ( x ) = m ⋅ x + n , m = 0.12 and n = -0.71 with 95% confidence intervals (0.026, 0.22) and (-
1202 0.91, -0.52) respectively and goodness of fit parameters, Sum Squared Error: 4.7109e-006,
1203 R-square: 0.8919, Root Mean Square Error: 7.6737e-004. In contrast the spontaneous
1204 retractions in the females has m = 0.10, n = -0.77 and confidence intervals (-0.01, 0.22) and

34
1205 (-1.00, -0.54) respectively. The goodness of fit parameters were SSE: 6.1033e-007, R-square:
1206 0.9033, Adjusted R-square: 0.8913, RMSE: 2.7621e-004. (B) The males of comparable age do
1207 not have distinct patterns of variability between the forward motions intended to the target
1208 and the spontaneous withdrawing segments of the reach. m = 0.57, n = -1.07, with 95%
1209 confidence intervals (-1.20, -0.94) and (0.24, 0.91) respectively and SSE: 7.9535e-007, R-
1210 square: 0.9715, RMSE: 2.8202e-004. In the withdrawing motions, m = 0.7342, n = -1.14 (-
1211 1.26, -1.01), (0.26, 1.20), SSE: 2.3818e-007, R-square: 0.9688, Adjusted R-square: 0.9657,
1212 RMSE: 1.5433e-004. Notice as well the differences in the slope and intercept of the linear
1213 regression fit that the power relation revealed on the log-log plot of the scatter (details
1214 reported in the main text). Each point of the scatter represents the empirically estimated
1215 scale and shape parameters of the two-parameter continuous Gamma family of probability
1216 distributions of one subject. The empirical frequency histograms were fit using MLE with
1217 95% confidence intervals. (C-D) Differences in the dispersion of the distributions quantified
1218 between typical females and males of college age. Females were more variable and spanned
1219 a broader range of values whereas males were less variable but spanned a narrower range
1220 of parameter values across the group. Each curve is the experimentally estimated Gamma
1221 pdf according to the empirically determined range of values for each person.
1222

1223

1224 Figure 4
1225 Patterns of variability from the percent of time to the maximum speed differed
1226 between TD participants and participants with ASD for each sex. (A-C) Each row is the
35
1227 principal components (PC) from the singular value decomposition of the covariance matrix
1228 (zero-mean centered) for each participant (28 females and 43 males). Forward segments of
1229 the reach aimed at the target did not reveal differences in the PC of the variability of this
1230 (kinesthetic) latency parameter. (B-D) The uninstructed retractions of the reach
1231 (withdrawing from the target) self-segregated two clusters within the TD group and one
1232 cluster in the ASD group for each sex. (C) The PC of TD females (below 5 years of age)
1233 grouped apart from older TD females (older than 5 years of age). Yet all females with ASD
1234 had PC patterns in the range of the young TD cluster. These included verbal and non-verbal
1235 females with ASD. (D) As in the young TD females, the young TD males clustered apart from
1236 the older TD males. And as with the females with ASD, the variability patterns of the timing
1237 of the maximum speed in the males with ASD grouped along the PC of the young TD. There
1238 were 2 outliers in the TD adults whose PC patterns fell on the side of the young TD (or ASD).
1239 Both were child prodigies (one had received a diagnosis of ASD by 3 years of age while the
1240 other never received an official diagnosis). Both are presently theoretical physicists (their
1241 cases will be discussed in detail elsewhere).
1242

1243

1244 Figure 5
1245 Movement decision latency across self-emerging clusters. (A) Frequency histograms of
1246 the movement decision time. Clusters are from self-emerging aggregates of the
1247 somatosensory velocity-dependent kinematics parameters. (B) Stochastic signatures of this
1248 parameter on the (a, b) Gamma plane. Note that young individuals with ASD (age 4-8) are
1249 closer to TD individuals below 4 years of age than to TD individuals older than 4 years of
1250 age, even though the latter TD group is closer in age than the former TD group.
1251

36
1252

1253 Figure 6
1254 Sex differences in the latency of the decision movement time (s) across self-emerging
1255 clusters. Color codes are as before (green TD < 4 years old; blue TD > 4 years old; red TD
1256 college level; magenta non verbal ASD 4-6 years old; maize non-verbal ASD 8-16 years old;
1257 black verbal ASD 10-15 years old). (A) Male participants overall show significant differences
1258 between ASD and TD (Kruskal-Wallis p<1.02x10-7, χ2 28.32). (B) The differences between
1259 females with ASD and TD are also significant (Kruskal-Wallis p<0.004, χ2 8.25), yet they were
1260 more pronounced in the younger girls around the age of 3-4 (TD) vs. 4-5 (ASD). By then, the
1261 differences in movement decision latency are significant in females (Kruskal-Wallis
1262 p<1.10x10-16, χ2 68.78) but not in males (Kruskal-Wallis p<0.32, χ2 1.01). Black arrows
1263 highlight the separation between males (A) and females (B) in the clusters of TD
1264 kindergarten (green) and 4-6 years old non-verbal participants with ASD. (C) Males and
1265 females with ASD showed significant differences across ages (Kruskal-Wallis p<4.3x10-9, χ2
1266 34.47);(Kruskal-Wallis p<1.0x10-14, χ2 59.87);(Kruskal-Wallis p<5.6x10-5, χ2 16.23). (D) TD
1267 males and females showed significant differences at the kindergarten (Kruskal-Wallis
1268 p<0.0009, χ2 10.98) and pre-school (Kruskal-Wallis p<0.0004, χ2 12.37) levels, but these
1269 were not significant at the college level (Kruskal-Wallis p<0.16, χ2 2).
1270

37
1271

1272 Appendix Figure


1273 Two variants of the pointing task to examine goal-directed and incidentally
1274 uninstructed movements in closed loop with decision-making processes. (A) Variant 1:
1275 basic pointing task, measuring the goal-directed motion to touch a target on the touch
1276 screen and spontaneous retraction away from it. The arrows mark the flow of motion. The
1277 speed profiles are also plotted as insets, with the dot marking the peak velocity and the
1278 arrow marking the time at which the peak is attained. (B) Variant 2: Decision making
1279 pointing during a match to sample task (upper-left and upper-right corners) matches the
1280 sample (bottom-center). A representative speed profile is also plotted as insets with the
1281 landmarks used to navigate the behavior. The touch at the bottom-center of the touch-
1282 screen simultaneously presents the sample and two possible targets. After the decision has
1283 been made, the hand goes to the targeted choice and touches the screen again. Examples of
1284 other stimuli – of varied cognitive load - used in the match-to-sample task are shown on the
1285 upper right hand corner. Notice that the basic pointing motion is the same in all cases, yet
1286 the changes in the complexity of the visual stimulus is used to probe the changes in the
1287 stochastic signatures of these motions continuously measured over time.
1288

38
1289 Tables

1290 Table 1 Scores, clinical assessments of the participants with ASD


Stanford-Binet ADOS Scores GARS Scores
Age Com + Stereo Com Soc Autism
Code Sex NVIQ VIQ FSIQ Stereo Com Soc
(yrs) Soc SS SS SS Index
01 M 4.3 42 43 40 4 8 13 21 N/A N/A N/A N/A
02 F 5.9 44 51 45 2 4 13 17 N/A N/A N/A N/A
03 M 6.0 N/A N/A 100 N/A N/A N/A N/A N/A N/A N/A N/A
04 M 6.3 N/A N/A 50 N/A N/A N/A N/A N/A N/A N/A N/A
05 M 7.6 50 46 45 3 6 11 17 N/A N/A N/A N/A
06 F 7.8 42 43 40 4 7 12 19 N/A N/A N/A N/A
07 M 7.8 42 43 40 2 6 14 20 N/A N/A N/A N/A
08 M 9.0 42 44 40 1 5 10 15 N/A N/A N/A N/A
09 M 9.9 42 43 40 4 5 8 13 N/A N/A N/A N/A
10 M 10 N/A N/A 107 3 3 9 12 N/A N/A N/A N/A
11 M 10.3 42 43 40 3 4 10 14 N/A N/A N/A N/A
12 M 11.5 100 82 90 7 5 6 11 N/A N/A N/A N/A
13 F 11.5 50 43 44 N/A N/A N/A N/A N/A N/A N/A N/A
14 M 11.7 42 43 40 5 8 10 18 N/A N/A N/A N/A
15 M 11.7 43 43 40 N/A N/A N/A N/A N/A N/A N/A N/A
16 M 12 N/A N/A 67 4 5 13 18 N/A N/A N/A N/A
17 F 12 N/A N/A 60 4 8 10 18 N/A N/A N/A N/A
18 M 12 N/A N/A 95 2 5 8 13 N/A N/A N/A N/A
19 M 12 N/A N/A 95 1 5 7 12 N/A N/A N/A N/A
20 M 13 N/A N/A 89 2 3 7 10 N/A N/A N/A N/A
21 M 13.8 42 43 40 N/A N/A N/A N/A N/A N/A N/A N/A
22 M 14 N/A N/A 74 3 9 10 19 N/A N/A N/A N/A
23 F 14.3 50 43 44 N/A N/A N/A N/A 8 11 9 124
24 F 15 N/A N/A 52 2 6 11 17 N/A N/A N/A N/A
25 F 15 N/A N/A 77 N/A N/A N/A N/A N/A N/A N/A N/A
26 F 15 N/A N/A 71 6 5 7 12 N/A N/A N/A N/A
27 M 15 N/A N/A 56 3 4 10 14 N/A N/A N/A N/A
28 F 15.8 42 43 40 N/A N/A N/A N/A 13 10 11 109
29 M 16 N/A N/A 100 N/A N/A N/A N/A N/A N/A N/A N/A
30 F 16 N/A N/A 81 2 7 9 16 N/A N/A N/A N/A
31 M 18 N/A N/A 101 2 4 6 10 N/A N/A N/A N/A
32 M 18 N/A N/A 96 4 4 8 12 N/A N/A N/A N/A
33 M 18 N/A N/A 76 1 5 7 12 N/A N/A N/A N/A
34 M 25 N/A N/A 99 6 3 7 10 N/A N/A N/A N/A

1291
1292 The first column identifies the self-emerging cluster number in Figure 3B of the main text (O is for
1293 outliers). Stanford-Binet 5th edition was used to assess intelligence of each target student with ASD
1294 (Roid 2003). A score of 100 corresponds to “normal” intelligence. A departure from 100 by 15 points
1295 indicates one standard deviation above or below normal intelligence. NVIQ is a measure of nonverbal

39
1296 IQ. VIQ is a measure of Verbal IQ. FSIQ is the sum of verbal and non-verbal intelligence scores
1297 converted to a standardized score. The ADOS (Autism Diagnostic Observational Scale; (Gotham et al.
1298 2009; Lord et al. 2000) is a standard assessment tool used to provide clinicians with a basis for the
1299 diagnosis of ASD. Module 1 of the ADOS was used for the young, non-verbal students. Module 3 was
1300 used for the adolescent students with conversation ability. Stereo is a measure of stereotyped
1301 behaviors on a score, with a higher score indicating more stereotyped behaviors; however there is no
1302 cutoff for a diagnosis of autism. Com is the total Communication score. A score of 4 is the cutoff for
1303 Autism, with a score of 2 being the cutoff for Autism Spectrum. Soc is the total Reciprocal Social
1304 Interaction Score. A score of 4 is the cutoff for Autism, with a score of 2 being the cutoff for Autism
1305 Spectrum. Com + Soc is the combined Communication and Social Interaction score, with a score of 12
1306 being the Autism cutoff, and a score of 7 being the Autism spectrum cutoff. Because of their age and
1307 extremely limited verbal ability, 2 children could not be given the ADOS. Therefore the GARS 2
1308 (Gilliam Autism Rating Scale – Second edition, (Gilliam 2006) was used to assess these individuals.
1309 Stereo SS is the standardized score of stereotyped behaviors. Com SS is the standardized score of
1310 Communication. Social SS is the standardized score of Social Interaction. The Autism Index is the sum
1311 of standard scores, converted to normed index score.

1312

40
1313 Table 2 Information from TD participants
Participants in biomechanical
Participants in both experiments
experiment only
Cluster ID Sex Age (Yrs) ID Sex Age (Yrs)
1 1 M 3.3 1 M 19
1 2 M 3.3 2 M 20
1 3 F 3.3 3 M 20
1 4 F 3.8 4 F 20
1 5 M 3.8 5 M 20
1 6 M 3.9 6 F 21
2 7 F 4.0 7 M 21
2 8 F 4.0 8 M 21
2 9 F 4.0 9 M 21
2 10 M 4.0 10 F 21
2 11 M 4.8 11 F 22
2 12 F 5.0 12 F 22
2 13 M 5.0 13 F 22
3 14 F 21 14 M 23
3 15 F 21 15 M 23
3 16 M 13* 16 M 23
3 17 M 22 17 F 23
3 18 F 23 18 F 23
3 19 M 23 19 F 24
3 20 F 26 20 F 25
3 21 F 26 21 M 60
3 22 M 30 22 M 61*
1314
1315 TD children represented by clusters 1 (green) and 2 (blue). Cluster 3 (red) consists of

1316 college-level participants. The 22 additional college-level participants who only participated

1317 in the biomechanical task are in the last 3 columns. (*) Two child prodigies who undertook

1318 advanced Math college classes at a very young age.

41

You might also like