Individuals with autism spectrum disorder have altered visual encoding capacity

Jean-Paul Noel; Ling-Qi Zhang; Alan A Stocker; Dora E Angelaki

doi:10.1371/journal.pbio.3001215

. 2021 May 12;19(5):e3001215. doi: 10.1371/journal.pbio.3001215

Individuals with autism spectrum disorder have altered visual encoding capacity

Jean-Paul Noel ^1,^‡, Ling-Qi Zhang ^2,^‡, Alan A Stocker ^2,^3,^‡, Dora E Angelaki ^1,^‡,^*

Editor: Adam Kohn⁴

PMCID: PMC8143398 PMID: 33979326

Abstract

Perceptual anomalies in individuals with autism spectrum disorder (ASD) have been attributed to an imbalance in weighting incoming sensory evidence with prior knowledge when interpreting sensory information. Here, we show that sensory encoding and how it adapts to changing stimulus statistics during feedback also characteristically differs between neurotypical and ASD groups. In a visual orientation estimation task, we extracted the accuracy of sensory encoding from psychophysical data by using an information theoretic measure. Initially, sensory representations in both groups reflected the statistics of visual orientations in natural scenes, but encoding capacity was overall lower in the ASD group. Exposure to an artificial (i.e., uniform) distribution of visual orientations coupled with performance feedback altered the sensory representations of the neurotypical group toward the novel experimental statistics, while also increasing their total encoding capacity. In contrast, neither total encoding capacity nor its allocation significantly changed in the ASD group. Across both groups, the degree of adaptation was correlated with participants’ initial encoding capacity. These findings highlight substantial deficits in sensory encoding—independent from and potentially in addition to deficits in decoding—in individuals with ASD.

It is increasingly recognized that individuals with Autism Spectrum Disorder (ASD) show anomalies in perception, and these have been recently attributed to altered decoding (i.e. interpretation of sensory signals). This study reveals that independent of these changes, individuals with ASD show upstream deficits in sensory encoding (i.e., how samples are drawn from the environment).

Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental condition with high prevalence [1] which ubiquitously afflicting brain function, from perception to (social) cognition. Recent reports have attempted to formalize the functional implications of the disorder within the language of statistical inference, proposing that across domains ASD is largely characterized by an imbalance in weighting incoming sensory evidence with prior knowledge about the world [2–9].

This formulation places the study of ASD within the encoding–decoding framework of perception (Fig 1A). Namely, sensory stimuli, θ, are first encoded by a noisy neural measure—i.e., sensory information is mapped onto bandwidth-limited neural responses. Then, for example, within the framework of statistical (Bayesian) inference, these measurements, m, inform an observer about the identity or magnitude of a stimulus variable (likelihood). Combined with prior expectations about the environment, p(θ), this results in a posterior belief p(θ|m). An estimate of the stimulus variable $\hat{θ} (m)$ is generated based on the posterior and the utility function participants employ for the task. Together, these components and their appropriate combination (i.e., combination of a likelihood with a prior resulting in a posterior that is used in accordance with a utility function to generate responses) represent the “decoding” of sensory representation. Anomalies in this decoding process have been the main focus of recent computational studies of ASD (e.g., imbalance in the weighting of likelihood versus priors; see citations above), with researchers recently also emphasizing a slow or inflexible updating of priors in ASD [10,11].

The optimal combination of likelihoods with priors is, however, only one of many potential decoding mechanisms. Other commonly studied decoders in perceptual science include maximum likelihood estimation (MLE; [12]), linear decoders (e.g., [13]), and in more recent years, decoders based on neural networks [14]. Given the array of potential decoders the brain may employ and the various ways in which these may malfunction in ASD [8], here we were instead interested in identifying potential deficits in the sensory encoding, and the updating of this encoding with sensory exposure and feedback, irrespective of differences in decoding. This exclusive emphasis on sensory encoding is seldom, if ever, explored within the ASD literature and does not detract from reported anomalies in sensory decoding. Rather, it provides the opportunity to cleanly disentangle and localize deficits in encoding from deficits in decoding.

As experimenters, however, we can only measure participants’ responses as reported by verbal or motor output. These reports are influenced by both encoding and decoding processes (including participants’ prior belief and utility function, or “response model”), thus rendering the characterization of a singular process (i.e., encoding or decoding) difficult. To address this challenge, and to specifically test the hypothesis that sensory encoding and its updating are aberrant in ASD, we leverage a principled framework for data analysis emphasizing sensory encoding independently from decoding (Fig 1B and 1C). Specifically, we compute Fisher information (FI), a measure of how much information participants’ internal representations carry about the physical stimulus presented, by using a lawful relationship between estimation bias, variance, and FI, known as the Cramer–Rao lower bound (Eq 1; [15,16]):

\sqrt{I_{F} (θ)} \geq \frac{[1 + b^{'} (θ)]}{σ (θ)} .

(Eq 1)

In other words, by measuring bias (b) and variance (σ²) in estimates $(\hat{θ})$ of θ, we can determine a lower bound on FI, i.e., the accuracy with which the stimulus variable is represented. Equating this bound with a direct index of sensory encoding (i.e., in effect replacing the greater than or equal sign with an equality in Eq 1) requires a single assumption—that the translation from sensory representations to behavioral outputs was uncorrupted by stimulus-dependent noise. Note that this assumption is (often implicitly) made by virtually all decoders in perceptual science (e.g., Bayesian decoders, MLE), and neural decoders (e.g., linear classifiers, population vector, etc.). The exact form of the decoder (i.e., idiosyncrasies in the mapping from sensory estimates to motor outputs) does not matter beyond this single assumption, given that any decoder will trade off bias and variance. We further motivate and demonstrate the validity of our approach via both a numerical simulation (Fig 1) and an analytical derivation (see S1 Text) showing that: (1) the same encoding process (Fig 1D) can produce very distinct patterns of estimation bias and variance given different decoders (Fig 1E), yet critically; (2) with the Cramer–Rao lower bound, we are able to precisely recover the encoding scheme in terms of FI, regardless of the differences in decoding (Fig 1F).

To identify anomalies in the capacity, and the updating of sensory encoding in ASD, we took advantage of the well-known “oblique effect” [17,18]. That is, humans demonstrate greater sensitivity and repulsive (i.e., away from) biases in perceiving gratings and motion around cardinal compared to oblique orientations [19,20]. This effect is thought to reflect the fine-tuning of sensory encoding to environmental statistics—with vertical and horizontal orientations being most common in nature [21,22]. Recent studies [16,23] further highlight that it is exactly this anisotropic property of sensory encoding that leads to the observed repulsive bias. In other words, the oblique effect is thought to reflect a property of sensory encoding, and not decoding. A standard Bayesian decoder with homogenous encoding cannot account for the oblique effect, given the bias is away from and not toward the prior [23]. Lastly, early studies demonstrated that orientation discrimination is readily impacted by perceptual learning, adaptation, and feedback (e.g., [24–26]) on a relatively short timescale. In turn, orientation estimation is an appropriate paradigm to study the updating and flexibility of the encoding process.

First, we had neurotypical and ASD individuals estimate orientations in the absence of feedback, in order to index their baseline accuracy in encoding individual stimulus orientations and overall encoding capacity. Then, to parallel observations emphasizing the slow updating of Bayesian priors in ASD [10,11] and establish whether this phenotype would extend to sensory encoding, we presented participants with an experimental distribution of orientations (i.e., uniform) that differs from the natural distribution (i.e., peaking at cardinal orientations) and provided participants with performance feedback. Feedback was given in order to accelerate any putative reallocation of encoding resources due to the change in orientation distribution.

Results

Heightened variability and reduced learning in autism

We asked participants to perform a visual orientation estimation task, first without feedback, and then in the presence of trial-by-trial feedback in an attempt to induce learning and updating of sensory encoding. Groups of ASD (n = 17) and neurotypical (n = 25) individuals matched via button press a briefly (120 ms) presented visual target of random orientation. Target orientations were drawn from a uniform distribution. Initially, no feedback was presented (woFB block), but in the second and third block of trials (wFB1 and wFB2; 200 trials/block), feedback was presented by overlaying the participant’s response and the target (Fig 2A, see Materials and methods for details). As shown for an example neurotypical (Fig 2B) and ASD (Fig 2C) individual, orientation perception was biased away from cardinal orientations in both groups—namely, both groups demonstrated an “oblique effect” [17,18]. With feedback, the bias seemingly dissipated in the neurotypical participant but not the ASD individual (Fig 2B and 2C scatter plot are individual responses, curve is a running average within a sliding 18° window, and vertical dashed lines are orientations at 45°, 90°, and 135°. See S1 and S2 Figs for similar plots for all individual participants).

Fig 2 — **(A)** A target orientation (Gabor) is briefly presented, and participants report their percept by orienting a line indicator (white) via left or right button press. No feedback is given on the first block of trials but is in subsequent blocks by overlaying the target orientation and the participant’s response. (B) Target orientations (x-axis) are drawn from a uniform distribution (individual dots are single trials). Y-axis indicates the bias for an example neurotypical participant, and lines are the running average within a sliding window of 18°. Different columns (and the color gradient) respectively show performance on the block without feedback (woFB; leftmost), on the first block with feedback (wFB1; center), and the second block with feedback (wFB2; rightmost). (C) follows the format in (B) while depicting the performance of an example ASD participant. Raw data and code underlying this figure can be found at **S1 Code**, and numerical values that make up this figure can be found at **S2 Data**. ASD, autism spectrum disorder.

These basic observations are also evident in group averages of neurotypical and ASD individuals (Fig 3). When presenting targets between 0° (horizontal) and 45°, the bias was on average positive (e.g., neurotypical group: 6.0° ± 0.26° (SE), p < 10⁻³), suggesting that horizontal gratings were perceived closer to the oblique 45°. Contrary, when presenting targets between 45° and 90° (vertical), the bias was negative (e.g., neurotypical group: −5.14° ± 0.26°, p < 10⁻³), suggesting that vertical gratings were again perceived closer to the oblique 45° (Fig 3A and 3B). Before feedback, there was no statistical difference in the overall magnitude (i.e., absolute value) of bias between neurotypical and ASD groups (neurotypical: 5.63° ± 0.12°; ASD: 5.99° ± 0.22°, Δ = 0.35° ± 0.25°, p = 0.154). On the other hand, when provided with feedback, orientation perception bias was reduced in the neurotypical group but much less so in the ASD group (reduction in average magnitude of bias between woFB and wFB2, neurotypical: Δ = 2.38° ± 0.17°, p < 10⁻³; ASD: Δ = 1.04° ± 0.31°, p < 10⁻³; ΔNeurotypical—ΔASD = 1.34° ± 0.35°, p < 10⁻³; Fig 3A and 3B, feedback conditions shown by a color gradient).

Regarding the variability of orientation perception, we found that at baseline (i.e., before feedback), the ASD group had larger variance in their estimates than the neurotypical group (neurotypical: 0.180 ± 0.006; ASD: 0.297 ± 0.010; Δ = 0.117 ± 0.11, p < 10⁻³; Fig 3C and 3D). This is in line with a growing literature suggesting heightened variability in ASD and the “neural unreliability thesis” of ASD [27–31]. Additionally, while variance was reduced with feedback in the neurotypical group, especially in the latter feedback block, this was not evident in the ASD cohort (reduction in overall variance between woFB and wFB2, neurotypical: Δ = 0.037 ± 0.008, p < 10⁻³; ASD: Δ = −0.012 ± 0.015, p = 0.414).

As expected, given these differences in bias and variability, the neurotypical group overall had better performance as measured by root-mean-square error (RMSE). This was true in the initial block of the experiment (neurotypical: 14.82° ± 0.35°, p < 10⁻³; ASD: 19.42° ± 0.47°, p < 10⁻³) and was exacerbated with feedback (Fig 3E and 3F). Namely, performance of the neurotypical, but not the ASD, group increased with feedback (reduction in overall RMSE between woFB and wFB2, neurotypical: Δ = 2.83° ± 0.48°, p < 10⁻³, ASD: Δ = −0.55° ± 0.71°, p = 0.78).

Reduced capacity and aberrant allocation of encoding resources in autism

Beyond characterizing the raw task performance of neurotypical and autistic individuals, our main goal here is to detail their encoding capacity, the allocation of these resources, and flexibility with which encoding resources are reallocated, given the artificial stimulus distribution presented in the experiment. Using the Cramer–Rao bound (Eq 1) we found that in both neurotypical and ASD individuals, FI peaked at cardinal orientations (Fig 4A). That is, initially the shape of FI (i.e., normalized $\sqrt{I_{F} (θ)}$ , see Materials and methods, Eq 3) in both groups qualitatively matched the previously measured distribution of orientations in natural images [21,22]. Additionally, since FI directly determines discrimination thresholds [16,32], the observed shape of FI also corroborates a heightened sensitivity to cardinal orientations [18] in both neurotypical and individuals with ASD. On the other hand, both total FI (i.e., $\int_{θ} \sqrt{I_{F} (θ)} d θ$ ) during the very first block of trials and how its magnitude changed over the course of the experiment differed for ASD and neurotypical groups. Already during the woFB block, total FI was significantly lower in ASD than neurotypical individuals (woFB, neurotypical: 14.67 ± 0.24, ASD: 11.28 ± 0.19, Δ = 3.39 ± 0.43, p < 10⁻³). Further, total FI increased from woFB to wFB2 for the neurotypical group (Δ = 1.96 ± 0.38, p < 10⁻³) but did not in the ASD group (Δ = −0.060 ± 0.27, p = 0.83, Fig 4B).

Fig 4 — **(A)** FI peaked at cardinal orientations for both neurotypical (blue) and ASD (red) individuals, similar to the natural scene statistics of orientations. Further, visual inspection suggests a flattening of this function with feedback in neurotypical participants (from dark to light blue) but not ASD participants (see panels C and E for quantification). Smoothing using a sliding window of 18° has been applied for visualization. (B) The total amount of FI was larger in neurotypical individuals than ASD at the outset and increased over the course of the experiment in neurotypical (blue color gradient) but not ASD (red color gradient) individuals. (C) The FI pattern as a function of orientation was quantified by 2 parameters, ω, which mixes a cardinal orientation prior with a uniform distribution as the normalized (square root of) FI, and λ, scaling total FI. (D) λ and (E) ω as a function of group (blue = neurotypical control, red = ASD) and block. Error bars are ± SEM across 5,000 bootstrap runs. Raw data and code underlying this figure can be found at **S1 Code**, and numerical values that make up this figure can be found at **S4 Data**. ASD, autism spectrum disorder; FI, Fisher information.

The above analyses suggest that similarly to their neurotypical counterparts, individuals with ASD have placed the bulk of their encoding resources in line with the distribution of orientations in the natural environment (i.e., at cardinal orientations). More broadly, this observation supports the efficient coding hypothesis [33], whereby encoding resources are characterized by FI and directly follow the shape of the stimulus distribution (see Materials and methods, Eq 2). In turn, to quantitatively assess encoding capacity and distribution of resources in both groups throughout the experiment, we parametrized the shape of FI as a weighted mixture of a “cardinal orientation distribution” and a uniform orientation distribution (with weights ω and 1−ω, respectively; Fig 4C; also see Materials and methods, Eq 4). A ω value of 0.5 approximates the previously measured prior distribution of orientation in natural images [21,22]. Presumably, allocation of encoding resources starts closer to the natural stimulus distribution for orientations (ω~0.5) and is then gradually molded to become more uniform (i.e., smaller ω). Thus, by parametrizing normalized $\sqrt{I_{F} (θ)}$ as a weighted sum of the cardinal orientation prior and the distribution of orientations presented during the experiment (i.e., uniform), we can quantify if and how feedback changed our participants’ allocation of encoding resources along a range, from the natural to the experimental stimulus distribution. Further, the overall magnitude of FI is determined by a scaling parameter, λ (Fig 4C, Eq 4 in Materials and methods). We fit the bias predicted by the parameterized FI to the measured bias given the measured variance (see Materials and methods, Eqs 5 and 6, and S1 and S2 Figs for the biases and fits of individual participants; see S3 and S4 Figs for parameters extracted for individual participants). Average R² for the neurotypical group: 0.770 ± 0.025; average R² for ASD group: 0.774 ± 0.034.

Examination of the fitted λ values confirmed that at the outset of the experiment, the ASD group had a lower total amount of FI compared to the neurotypical group (neurotypical versus ASD at woFB; Δ = 4.352 ± 0.473, p < 10⁻³; Fig 4D). This effect is unlikely due to differences in task strategies, such as speed–accuracy trade-offs, given that the distribution of response times (RTs) was qualitatively similar across both groups, and if anything, slower in ASD, making this latter group deficitary in both speed and accuracy (S5 Fig). Over the course of the experiment, λ increased further for the neurotypical group (woFB versus wFB2 in neurotypical individuals: Δ = 2.094 ± 0.580, p < 10⁻³; Fig 4D), but this was not the case for the ASD group (Δ = −0.150 ± 0.441, p = 0.732, Fig 4D). These results suggest a lower overall encoding capacity in ASD relative to their neurotypical counterparts and further a capacity that does not improve over the course of repeated feedback. On the other hand, we observed a significant decrease in median RT between woFB and wFB2 for both the neurotypical (ΔRT = −0.61 ± 0.069 sec, p < 10⁻³) and ASD group (ΔRT = −0.75 ± 0.114 sec, p < 10⁻³; see S5 Fig), again suggesting that differences in task strategy or putative differences in working memory are unlikely to account for the differing accuracy and precision in visual orientation estimation across ASD and neurotypical individuals.

Most interestingly, regarding participants’ allocation of encoding resources, or their prior distribution of orientation as derived from a parametrized normalized $\sqrt{I_{F} (θ)}$ , both groups started with a similar ω parameter before feedback (Δ = 0.052 ± 0.043, p = 0.230, Fig 4E). That is, these groups allocated resources similarly in the natural environment where horizontal and vertical edges are most common. When provided with feedback (woFB versus wFB2), the ω parameter decreased for the neurotypical group (change in ω: Δ = 0.191 ± 0.046, p < 10⁻³, Fig 4E, blue) but much less so for the ASD group (change in ω: Δ = 0.086 ± 0.049, p = 0.080). Again, under the efficient coding framework, this change in the ω parameter can be viewed as participants of the neurotypical group reallocating their encoding resources, from following the natural distribution of orientations to now incorporating the uniform stimulus distribution imposed by the experiment. The ASD group, instead, did not readily update their allocation of encoding resources (see S3 and S4 Figs for individual participant data).

Reduced initial encoding capacity is correlated with inflexibility in updating resource allocation

We further questioned whether the herein described anomalies in sensory encoding present in ASD may relate to their purported inflexibility in updating priors [10,11]. Indeed, recent theoretical studies [34,35] highlight an inherent relation between encoding resources and the degree to which one can adapt to changing environments. Intuitively, if resources are allocated primarily to represent statistically likely events, then fewer can be devoted to distinguishing between statistically unlikely alternatives—and it is precisely these latter ones that are most informative regarding a potential change in the environment [34,35]. To examine this hypothesis, we correlated participants’ encoding capacity (λ) before feedback with the shape of their prior after feedback (ω at wFB2). We found a strong correlation between these variables (R² = 0.407, p < 10⁻³, Fig 5A). The correlation holds separately for the neurotypical and ASD groups (Fig 5B). Thus, the inflexibility in resource reallocation observed in individuals with ASD may emanate from their limited encoding resources.

Fig 5 — **(A)** The flatness of FI after feedback (y-axis, indicating incorporation of stimulus statistics from the experiment) correlates strongly with total encoding resources before feedback (x-axis). The blue dots indicate individual neurotypical participants, and red dots indicate individual ASD participants. (B) The same correlation holds within both the neurotypical and ASD groups. Raw data and code underlying this figure can be found at **S1 Code**, and numerical values that make up this figure can be found at **S5 Data**. ASD, autism spectrum disorder; FI, Fisher information.

Discussion

By tracking perceptual estimates while (i) exposing observers to an artificial stimulus distribution (i.e., uniformly distributed orientations) and presenting them with performance feedback, (ii) directly extracting FI of sensory encoding from psychophysical data, and (iii) parameterizing the shape of FI as a mixture of the cardinal and uniform distributions of orientation stimulus, we have estimated the total sensory encoding capacity and the change in its allocation with feedback in individuals with ASD. This combined approach resulted in a number of empirical observations. First, the initial pattern of FI as a function of orientation strongly resembled the known natural distribution of orientations in the natural environment [21,22] in both ASD and neurotypical individuals. Second, already in the first block of trials, the total encoding resources for the representation of gratings in ASD is about 2/3 that of their neurotypical counterparts. Third, exposure to an experimental distribution of orientations, coupled with feedback, resulted in the appropriate reallocation and an overall increase in encoding resources of neurotypical individuals. This was not the case in individuals with ASD. Lastly, initial encoding capacity correlated with the degree of adaptation from the natural to experimental allocation of encoding resources.

The observation that encoding capacity differs from the outset between ASD and neurotypical individuals is in line with “neural unreliability thesis” [27] and a number of recent reports that have suggested heightened variability in ASD, be it in relatively low-level visual psychophysics [36,37], more naturalistic and continuous tasks [31], neurovascular [28,29] or physiological responses [27]. Particularly noteworthy and relevant to the current report is a study by Park and colleagues [37], where researchers had neurotypical and ASD individuals discriminate visual orientations. Then, via psychophysical modeling [38,39], they were able to ascertain that individuals with ASD had increased internal noise and a reduced capacity for external noise filtering than neurotypical counterparts. These results are also in line with accounts suggesting a reduced use of Bayesian priors [2] or an overweighting of sensory likelihoods versus expectations in ASD [5]. Namely, these expectations in essence serve as low-pass filters reducing trial-to-trial variability, and thus a decreased use of Bayesian priors could engender greater variability.

The second major observation is that the reallocation of sensory encoding resources is inflexible in ASD relative to neurotypical individuals. Given that within the efficient coding hypothesis the allocation of encoding resources may directly reflect participants’ prior in the Bayesian framework [23], these results are similarly consistent with recent experimental findings suggesting that individuals with ASD may update their priors slowly [11], putatively due to overestimating the volatility of the environment [10]. Further, here we suggest an explanation for the apparently inflexibility of priors in ASD. Namely, when resources are allocated primarily to represent statistically likely events, the representation of a given external environment may be best. However, it also limits the degree of resources that may be devoted to distinguishing between statistically unlikely alternatives [34,35]. These statistically unlikely events, however, are the most informative when attempting to infer whether the environment has changed. There is an inherent trade-off between coding for representational fidelity and for inference of changing environments given limited encoding resources [34,35]. Thus, the pattern of results suggesting reduced encoding resources in ASD may indicate that this deficit is at the root of aberrant perceptual flexibility in ASD [10,11]. This speculation is supported by a correlation between the total amount of encoding resources a participant demonstrated in the initial block of the experiment, and the degree to which their sensory encoding had adapted at the end of the experiment.

Related to the reallocation of sensory resources, studies on perceptual learning have identified a multitude of processes that may contribute to learning, even in simple tasks. These include changes in low-level representations, changes in attention, decision rules, and/or the assimilation of feedback [40,41]. Thus, one may wonder whether the observed inflexibility of sensory encoding in ASD reflects a more global learning deficit. We do not believe so, for the following reasons. First, the total amount of encoding resources differed between ASD and neurotypical individuals already at the outset of the experiment, and, as stated above, this quantity correlated with the degree of reallocation of encoding resources with feedback. Thus, if the inflexibility of encoding was expressing a global deficit in incorporating feedback, one would have to account for its correlation with total encoding resources at outset. Second, individuals with ASD did learn. Their allocation of encoding resources drifted from the natural toward the experimental distribution, particularly in wFB2 (S6 Fig), just not as readily as in neurotypical individuals. More importantly, their response times decreased over the course of the experiment, and this effect was in fact “larger” in ASD than neurotypical individuals (S5 Fig). Similarly, Harris and colleagues (2015) [42] have reported normal perceptual learning in individuals with ASD, in terms of improvement in discrimination threshold, as long as training and testing samples completely overlap (as was the case in the current experiment). It is only when probing for generalization in perceptual learning (i.e., from one retinotopic location to another) that poor performance is observed in ASD [42], again demonstrating a more specific, rather than a general, form of learning deficit.

There is a wide range of potential neural mechanisms by which the amount and allocation of encoding resources may change on a timescale that is compatible with the duration of our experiment. While neural recordings in animal models have shown that practicing orientation discrimination can reshape tuning functions as early as V1 [43], given the timeframe of the current experiment, we consider that network-level mechanisms are most likely. A first possibility is a gain modulation of the neural response. Indeed, for an efficient coding framework where each cell transmits an equal portion of the stimulus probability mass, an increase in the overall firing rate corresponds to a direct increase in population FI [44]. Previous studies of sensory adaption (e.g., [26,45]) have also suggested that gain changes specific to a subpopulation of neurons are able to alter the allocation of coding resources. Gain modulation could be implemented by alterations in neuromodulation, and fittingly, Lawson and colleagues (2017) [10] recently demonstrated abnormal noradrenergic responsivity in ASD and suggested that this anomaly is related to the slow updating of their priors. The present findings suggest a functional interpretation for such a link: Abnormal noradrenergic responsivity could decrease FI (e.g., encoding resources), which then directly correlates with an inflexibility in altering priors. A second potential mechanism may be changes in the noise-correlation structure in the neural population. For example, a decrease in interneuronal correlations generally reduces the impact of noise on stimulus representation, thus leading to a higher population FI [46]. Along this line, Coen-Gagli and Solomon (2019) [47] have recently suggested that divisive normalization [48] is a critical player in neural variability, and Rosenberg and colleagues (2015) [49] have accounted for a wide array of perceptual deficits in ASD by postulating a deficit in divisive normalization.

For all these seemingly converging lines of evidence, however, there are open empirical questions.

First, in the current experiment, we provided participants with feedback on their performance in latter blocks of trials. This was done to accelerate learning of the novel statistical distribution of orientations and thus to enable reallocation of encoding resources. And indeed, scrutiny within the woFB block suggests that no learning occurred—neither in neurotypical nor ASD individuals—in the absence of feedback (S6 Fig), at least in the timeframe of our experiment. Thus, in the context of this experiment, seemingly feedback was necessary in reshaping encoding resources. This speculation goes hand in hand with our conjectures regarding the neural implementation of the current effect—likely being a network-level effect on the (feedforward) sensory representations in early visual areas (see [50] for an argument that the distribution of tuning functions in V1 underlies the oblique effect). In follow-up experiments, it will be interesting to repeat the current paradigm while considerably expanding the exposure time to the novel orientation distributions and without presenting feedback, in order to further understand if and how the updating of sensory encoding differs with mere exposure and with feedback.

Second, one may wonder about the potential impact of additional stimulus-independent variability introduced by factors following encoding—such as an inefficient decoder (i.e., a decoder that produces larger variance than that predicted by the Cramer–Rao lower bound), or post-perceptual factors such as motor noise. We attempted to experimentally minimize the potential impact of motor noise in ASD [51] by making the estimation task an unspeeded binary button press (i.e., up or down) that additionally had to be “confirmed” via another button press (see Materials and methods). Likewise, scrutiny of response time distributions did not suggest different strategies across experimental groups. Further, even if there were differences in post-encoding processes, these differing strategies (part of the decoding process) trade off biases and variances, and thus we would regardless recuperate an appropriate estimate of encoding (see Fig 1 and S1 Text). Nevertheless, to further ascertain that potential stimuli-independent variability or post-perceptual factors do not deter from the current conclusions we ran additional simulations. We show that turning the decoder inefficient by adding homogeneous late noise will both decrease the total amount and flatten the pattern of extracted FI (S7 Fig). This is incompatible with the observed results in the following aspects. First, our empirical data suggest exposure to an artificial stimulus distribution coupled with feedback increases total FI (i.e., the λ parameter) while flattening the pattern of FI (i.e., the ω parameter). Thus, if the increase in total FI with feedback in the neurotypical group were to be explained by an increase in the efficiency in the decoder or a decrease in motor variability, the pattern of FI ought to be sharpened, which is the exact opposite of what we observed in the data. Similarly, if we were to explain the lack of change in total FI in the ASD group with increase in additional noise, the pattern of FI would have to go through even less learning (S8 Fig). Lastly, if we were to explain the initial difference in total FI (i.e., before feedback) between ASD and neurotypical individuals with additional stimulus-independent variability in the ASD group, then the “true” (i.e., uncorrupted by late noise) FI pattern in ASD would have to be more extreme (i.e., sharper at cardinal orientations) than what is specified by the stimulus distribution in the natural environment (see S8 Fig). While we cannot completely rule out this latter possibility, we consider that it is most parsimonious that both neurotypical and ASD groups do indeed start with a pattern of FI that is similarly adapted to the natural statistics of orientations, just as we observed in our experiment.

Third, here we have exclusively focused on sensory encoding in an effort to separately understand each piece leading to perception (i.e., encoding, recoding, decoding). However, comprehensively understanding sensory processing in ASD will ultimately necessitate a broader framework including encoding, prior beliefs, and loss (utility) functions, among others. In the context of oblique effect this will require a nontrivial form of encoding [23], orientation prior [21], and loss function [52]. Thus, instead of “reverse-engineering” all these components together [53,54], which would require a set of additional assumptions, here we took a data-driven approach. Our method, while not a complete model, can isolate and directly characterize the encoding process while being agnostic to potential differences in the decoding process and free of parametric restrictions about the shape of, for example, priors and likelihoods (see S1 Text for an extended discussion). As stated above, this is not to say that there are no differences in sensory decoding between the neurotypical and ASD group but to simply emphasize anomalies in encoding. Future work will be able to build on this foundation to extend the comparison of sensory processing in ASD relative to neurotypical individuals to a full model.

Lastly, we must mention that in a standard Bayesian framework, the lower encoding capacity of ASD would likely translate to broader likelihood functions or underweighting of sensory evidence. While some prior work [29,30,36,37] indeed suggest less precise sensory representation in ASD, others [55–58] have not replicated these observations. Butler and colleagues (2017) [58], for example, have shown no difference in the reliability of visual and somatosensory evoked activity during passive observation. Furthermore, some [3,54] have even argued for more, and not less, precise sensory likelihoods in ASD. In turn, future work will have to charter the domain generality of deficits and/or improvements in sensory encoding in ASD. More broadly, we suggest caution must be taken when comparing “likelihood functions” in an encoding–decoding framework across studies, given that these represent an abstract and aggregated property of a complex and multilayered hierarchical neural process, and could be strongly determined by factors such as stimulus type and task context. Here, therefore, instead of broadly describing potential anomalies in likelihood functions and/or prior distributions, we have taken a principled approach that links perceptual bias and variance to the available information in the encoding step. There are no degrees of freedom, and thus the problem is not underconstrained. Instead, FI is completely constrained by the measured pattern of biases and variance across the state space of interest (here 180 degrees of possible orientations). Further, by following the efficient coding hypothesis, we can relate encoding to perceptual priors directly from psychophysical data without explicit parametric assumptions. Having taken these properties into account, the present analyses suggest that in ASD a reduced pool of encoding resources leads to increased variability, and then putatively to the inability to accurately detect and represent statistically unlikely events. This may explain the observed inflexibility in the reallocation of sensory resources following a change in the environment [10,11]. More importantly, our results highlight that the encoding stage itself appears aberrant in ASD, and thus future studies examining perception in ASD must carefully characterize other aspects of normative computation beyond Bayesian decoding.

Materials and methods

All data and code for our analysis can be found at S1 Code or at https://github.com/cpc-lab-stocker/ASD_Encoding_2020

Participants

A total of 42 participants completed an orientation-matching task. Seventeen were individuals diagnosed as within the ASD (N = 17, mean ± SD; age = 15.3 ± 2.6 years; AQ = 33.5 ± 7.0; SCQ = 16.8 ± 4.4) by expert clinicians. The rest were neurotypical individuals (N = 25, mean ± SD; age = 14.8 ± 2.1 years; AQ = 14.0 ± 5.5; SCQ = 5.6 ± 3.2). Participants had normal or corrected-to-normal vision and no history of musculoskeletal or neurological disorders. Before partaking in the study, all participants completed the Autism Spectrum Quotient (AQ; [59]) and the Social Communication Questionnaire (SCQ; [60]). The Institutional Review Board at Baylor College and Medicine approved this study (protocol number H-29411) in adherence to the Declaration of Helsinki, and all participants gave their written informed consent and/or assent.

Materials and procedures

Participants were comfortably seated facing a gamma-corrected CRT monitor (Micron Technology, Boise, Idaho; 43 × 35 cm) at a distance of 57 cm. Participants self-initiated a trial by button press, upon which a Gabor (120 ms presentation, 0.4 cycles/degree, 10 degrees, Gaussian envelope) was presented centrally on a gray background. The orientation of this target Gabor was random (uniform distribution, 0 to 180 degrees). Immediately following the offset of the Gabor, a mask consistent of 6 concentric circles (line color: black; 2.5, 4, 5.5, 7, 8.5, and 10 degrees) was presented. This mask had a duration of 300 ms and was presented in order to eliminate the possibility of participants experiencing an afterimage. Following a blank period of 400 ms, participants were presented with a white Gabor patch (3 cycles/degree, only 1 strip visible, random initial orientation) that they rotated via button press (up and down arrow, resolution = 1 degree) until they considered the orientation of the white Gabor patch indicator to match that of the target Gabor. Participants logged a response via a “confirmatory” button press. The intertrial interval was set to 1 second, and participants completed 200 trials per block. The experiment consisted of 3 blocks; the first was without feedback, as described above. In the second and third blocks, participants were given feedback by superimposing the target Gabor and the orientation reported during the intertrial interval. Participants were given approximately 5-minute rest between blocks. All stimuli were generated and rendered using C++ Open Graphics Library (OpenGL).

Data analysis

We model participants as a generic estimator T of the orientation parameter θ, and thus the responses as independent samples of T. The bias and variance of the estimator T is a function of θ defined by b(θ) = E_θ[T]−θ, and σ²(θ) = σ²_θ[T], respectively. A sliding window (of size 18° when analyzing individual participants and 4° for the combined participant) is used to calculate the function b(θ) and σ²(θ). Within each window, due to the circular nature of orientation, the mean (μ) and concentration (κ) parameters are calculated by fitting a von Mises distribution. The bias b was then defined as the difference between μ and the center of the window. The variance σ² is defined as σ² = 1/κ, and standard deviation $σ = \sqrt{σ^{2}}$ .

The lower bound of FI as a function of θ is given by Cramer–Rao lower bound (Fig 1):

\sqrt{I_{F} (θ)} \geq \frac{[1 + b^{'} (θ)]}{σ (θ)} .

(Eq 1, see main text)

For our analysis, we assumed a tight bound (e.g., a wide range of commonly used decoders, including Bayesian and maximum likelihood estimators can attain the bound, also see Fig 1 and S1 Text), which allows us to extract FI directly from data. Furthermore, with an efficient coding assumption that the square-root of FI is directly proportional to the prior distribution of the stimulus:

\sqrt{I_{F} (θ)} \sim p (θ) .

(Eq 2)

We can thus estimate the prior distribution that corresponds to an efficient sensory representation directly from data by calculating the normalized square-root of FI (Fig 4A):

\sqrt{I_{F} (θ)} / \int_{θ} \sqrt{I_{F} (θ)} d θ .

(Eq 3)

Note that the denominator, $\int_{θ} \sqrt{I_{F} (θ)} d θ$ , is also a direct measure of the total amount of resource available for encoding (Fig 4B; [16,23]).

Next, to quantify the changes in encoding during the experiment, we parameterized the square-root of FI with λ and ѡ:

\sqrt{I_{F}} (θ; λ, ѡ) = λ * (ѡ * [0.877 * (1 - s i n | 2 θ |)] + (1 - ѡ) * \frac{1}{π}),

(Eq 4)

where λ determines the amount of total (square-root of) FI, and ѡ controls the shape of FI (i.e., allocation of encoding resources, effectively the prior distribution under the efficient coding hypothesis) by mixing a “cardinal orientation prior” with a uniform orientation distribution (Fig 4C). Note the constants are such that the prior term of the equation normalizes properly to 1.

To recover λ and ѡ, we calculate the predicted bias:

\hat{b} (θ; λ, ѡ) = \int_{0}^{θ} (\sqrt{I_{F}} (θ; λ, ѡ) * σ (θ) - 1) d θ .

(Eq 5)

We find $\hat{λ}$ and $\hat{ѡ}$ that give rise to the best fit to the observed bias b(θ) using MATLAB’s fmincon:

{a r g m i n}_{λ, ѡ} {| | \hat{b} (θ; λ, ѡ) - b (θ) | |}_{2}^{2} .

(Eq 6)

See S1 and S2 Figs for individual participant fits. See S3 and S4 Figs for parameters extracted for each individual participant and their relationship with measures of autism symptomatology. Note that in principle, we can fit $\sqrt{I_{F}} (θ; λ, ѡ)$ to the extracted FI directly. Here, we choose to fit the bias pattern b(θ), to avoid the noisy derivative, b′(θ) in the Cramér–Rao lower bound, and effectively replaced it with an integration step instead.

Fig 4D and 4E show the parameters estimated for the combined participant, while the regression analysis presented in Fig 5 is based on parameters estimated with the same procedure applied to each individual participant. All statistical tests and p-values (except for the regression in Fig 5) are two-sided and are based on the distribution (intervals) of the sample statistics (e.g., mean, variance, model parameters ω and λ) across 5,000 bootstrap runs ([61]).

Supporting information

S1 Text. Motivation and analytical derivation for using the Cramer–Rao lower bound to measure Fisher information of the encoding process.

(DOCX)

Click here for additional data file.^{(40.4KB, docx)}

S1 Fig

Scatter plot of the response data for individual neurotypical participants (target on x-axis and bias on y-axis), and the model fits (dotted lines) to the average bias (solid lines). The number on the top right of each panel indicates the goodness-of-fit (R²). Average R²: 0.770±0.025. Note that we enforced the bias pattern to have a periodicity of 90°. Color gradient (from dark to light) and left-to-right shows first the block without feedback (woFB), and then the first and second blocks with feedback (wFB1, wFB2). Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S6 Data.

(EPS)

Click here for additional data file.^{(17.7MB, eps)}

S2 Fig

Scatter plot of the response data for individual ASD participants (target on x-axis and bias on y-axis), and the model fits (dotted line) to the average bias (solid line). The number on the top right of each panel indicates the goodness-of-fit (R²). Average R²: 0.774±0.034. Note that we enforced the bias pattern to have a periodicity of 90°. Color gradient (from dark to light) and left-to-right shows first the block without feedback (woFB), and then the first and second blocks with feedback (wFB1, wFB2). Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S6 Data. ASD, autism spectrum disorder.

(EPS)

Click here for additional data file.^{(12.1MB, eps)}

S3 Fig

Magnitude λ (top) and shape ω (bottom) of FI as a function of AQ scores for individual participants. Neurotypical individuals are depicted in blue, and ASD participants are shown in red (individual dots are single participants, dots with errors bars are means across individuals ± SEM). The results corroborate the conclusion (Fig 4D and 4E) that the neurotypical and ASD group differed only by the λ parameter before feedback but both the λ and ω parameters after feedback. Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S5 Data. AQ, Autism Quotient; ASD, autism spectrum disorder; FI, Fisher information.

(EPS)

Click here for additional data file.^{(2MB, eps)}

S4 Fig

Magnitude λ (top) and shape ω (bottom) of FI as a function of SCQ scores. Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S5 Data. FI, Fisher information; SCQ, Social Communication Questionnaire.

(EPS)

Click here for additional data file.^{(2MB, eps)}

S5 Fig. Histogram of RT (i.e., time it took for participants to rotate the response Gabor patch to match the target orientation).

The overall shape of the distributions is qualitatively similar. There is a significant albeit small difference in the median RT between the neurotypical and ASD group across all 3 blocks (woFB, wFB1, and wFB2). Furthermore, there is a significant decrease in RT for both the neurotypical and ASD group with feedback as the experiment progresses. Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S6 Data. ASD, autism spectrum disorder; RT, response time.

(EPS)

Click here for additional data file.^{(2.5MB, eps)}

S6 Fig

We conducted a sliding-window analysis in an attempt to reveal the effect of learning within each block (woFB, wFB1, and wFB2) in terms of changes in: (A) average magnitude of the bias; (B) standard deviation of the response; (C) shape of the FI ω; and (D) magnitude of the FI λ. Our analysis suggested that, within the context of our experiment, feedback seems essential for learning. In fact, within the first block (woFB), the bias and variance of both groups seem to be increasing as opposed to decreasing. Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S6 Data. FI, Fisher information.

(EPS)

Click here for additional data file.^{(3.6MB, eps)}

S7 Fig

(A) We simulated the potential impact of stimulus-independent noise/variability introduced by factors other than encoding, such as an inefficient decoder (i.e., a decoder that produces larger variance than that dictated by the Cramer–Rao lower bound), or post-decisional factors such as motor noise. We consider this important, given that while we attempted to reduce motor responses to their simplest form (i.e., nonspeeded binary button press as opposed to more complex movements such as joystick use), the presence of increase motor noise is well documented in ASD (e.g., Gowen and Hamilton, 2012). (B) An increase in stimulus-independent, post-encoding noise will cause a decrease in the overall magnitude of the extracted FI. (C) An increase in post-encoding noise will flatten the shape of the extracted FI, while decrease the overall magnitude of FI. These results are incompatible with what we observed in our experiment, since we observe either an increase in the magnitude of FI associated with a flattening in the shape of FI (neurotypical group), or no change in the magnitude of FI, while the shape of FI is slightly flattened (ASD). Therefore, these simulations argue that the differences we herein report cannot be attributed to motor noise or other sources of noise that are stimuli-independent (see Discussion and S8 Fig). Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S6 Data. ASD, autism spectrum disorder; FI, Fisher information.

(EPS)

Click here for additional data file.^{(3MB, eps)}

S8 Fig. In a further attempt to understand the potential impact of post-encoding noise, we conducted model simulations with the counterfactual assumption that the abnormalities in λ (both total and change thereof) in the ASD group are fully explained by stimulus-independent noise.

(A) We simulated an artificial observer (black solid) for which its FI increased during wFB2, just as for neurotypical individuals. However, due to additional noise, this observer appears (black dashed) to be unchanged, similar as the ASD group (red dashed). (B) The ω values in the artificial observer (black solid) would have to undergo less learning than what was observed in ASD, such that that when it is corrupted by additional noise, it appears to have similar ω values as the ASD group (red dashed). (C) Similarly, we simulated another artificial observer (black solid) for which its FI matched that of the neurotypical individuals during woFB (blue dashed), yet appears to be reduced (black dashed) to a similar level as the ASD group (red dashed) due to additional noise. (D) The ω values in the artificial observer (black solid) would have to be even more extreme (FI peaking even stronger than what is determined by the natural distribution of orientations in the environment), such that when it is corrupted by additional noise, it appears to have similar ω values as the ASD group (red dashed). We consider this unlikely since it is most parsimonious that both neurotypical and ASD groups do indeed start with a pattern of FI that is similarly adapted to the natural statistics of orientations. Importantly, in both cases, the ω values in the simulated observer strongly support our main conclusion that the neurotypical group reallocate their encoding resources toward uniform distribution more so than the ASD group did. Raw data and code underlying this figure can be found at S1 Code, and numerical values that make up this figure can be found at S6 Data. ASD, autism spectrum disorder; FI, Fisher information.

(EPS)

Click here for additional data file.^{(2.8MB, eps)}

S1 Code. Raw data and MATLAB code for our analysis and figures.

Routine to reproduce Figs 1D–1F, 2B, 2C, 3, 4 and 5 and S1–S6, panel B in S7, and S8 Figs.

(ZIP)

Click here for additional data file.^{(1.1MB, zip)}

S1 Data. Numerical values that make up Fig 1D–1F.

(ZIP)

Click here for additional data file.^{(5.8MB, zip)}

S2 Data. Numerical values that make up Fig 2B and 2C.

(ZIP)

Click here for additional data file.^{(82.4KB, zip)}

S3 Data. Numerical values that make up Fig 3.

(ZIP)

Click here for additional data file.^{(19MB, zip)}

S4 Data. Numerical values that make up Fig 4.

(ZIP)

Click here for additional data file.^{(13.3MB, zip)}

S5 Data. Numerical values that make up Fig 5 and S3 and S4 Figs.

(ZIP)

Click here for additional data file.^{(16.7KB, zip)}

S6 Data. Numerical values that make up S1, S2, S5, S6, panel B in S7, and S8 Figs.

(ZIP)

Click here for additional data file.^{(6.6MB, zip)}

Acknowledgments

We thank J. Patterson and A. Rosenberg for piloting some of the stimuli used in the present experiments and H. Park for participation in data collection.

Abbreviations

AQ: Autism Spectrum Quotient
ASD: autism spectrum disorder
FI: Fisher information
MLE: maximum likelihood estimation
RMSE: root-mean-square error
RT: response time
SCQ: Social Communication Questionnaire

Data Availability

The data underlying each figure can be found in supplementary files. Additional data and code for our analysis can be found at: https://github.com/cpc-lab-stocker/ASD_Encoding_2020.

Funding Statement

This work was supported by the Simons Foundation, via a SFARI Grant (#396921) to D.E.A. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

1.Xu G, Strathearn L, Liu B, O’Brien M, Kopelman TG, Zhu J, et al. Prevalence and Treatment Patterns of Autism Spectrum Disorder in the United States, 2016. JAMA Pediatr. 2019. February 1;173(2):153–9. 10.1001/jamapediatrics.2018.4208 [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Pellicano E, Burr D. When the world becomes ’too real’: a Bayesian explanation of autistic perception. Trends Cogn Sci. 2012. October;16(10):504–10. 10.1016/j.tics.2012.08.009 [DOI] [PubMed] [Google Scholar]
3.Brock J. Alternative Bayesian accounts of autistic perception: comment on Pellicano and Burr. Trends Cogn Sci. 2012. December;16(12):573–4; author reply 574–5. 10.1016/j.tics.2012.10.005 [DOI] [PubMed] [Google Scholar]
4.Van de Cruys S, Evers K, Van der Hallen R, Van Eylen L, Boets B, de-Wit L, et al. Precise minds in uncertain worlds: predictive coding in autism. Psychol Rev. 2014. Oct;121(4):649–75. 10.1037/a0037665 [DOI] [PubMed] [Google Scholar]
5.Lawson RP, Rees G, Friston KJ. An aberrant precision account of autism. Front Hum Neurosci. 2014;8:302. 10.3389/fnhum.2014.00302 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Powell G, Meredith Z, McMillin R, Freeman TC. Bayesian models of individual differences: Combining autistic traits and sensory thresholds to predict motion perception. Psychol Sci. 2016. December;27(12):1562–72. 10.1177/0956797616665351 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Karaminis T, Cicchini GM, Neil L, Cappagli G, Aagten-Murphy D, Burr D, et al. Central tendency effects in time interval reproduction in autism. Sci Rep. 2016. June 28;6(1):1–3. 10.1038/s41598-016-0001-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Haker H, Schneebeli M, Stephan KE. Can Bayesian theories of autism spectrum disorder help improve clinical practice?. Front Psych. 2016. June 17;7:107. 10.3389/fpsyt.2016.00107 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Palmer CJ, Lawson RP, Hohwy J. Bayesian approaches to autism: Towards volatility, action, and behavior. Psychol Bull. 2017. May;143(5):521. 10.1037/bul0000097 [DOI] [PubMed] [Google Scholar]
10.Lawson RP, Mathys C, Rees G. Adults with autism overestimate the volatility of the sensory environment. Nat Neurosci. 2017. September;20(9):1293. 10.1038/nn.4615 [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Lieder I, Adam V, Frenkel O, Jaffe-Dax S, Sahani M, Ahissar M. Perceptual bias reveals slow-updating in autism and fast-forgetting in dyslexia. Nat Neurosci. 2019. February;22(2):256–64. 10.1038/s41593-018-0308-9 [DOI] [PubMed] [Google Scholar]
12.Ernst MO, Banks MS. Humans integrate visual and haptic information in a statistically optimal fashion. Nature. 2002. January;415(6870):429–33. 10.1038/415429a [DOI] [PubMed] [Google Scholar]
13.Hung CP, Kreiman G, Poggio T, DiCarlo JJ. Fast readout of object identity from macaque inferior temporal cortex. Science. 2005. November 4;310(5749):863–6. 10.1126/science.1117593 [DOI] [PubMed] [Google Scholar]
14.Walker EY, Cotton RJ, Ma WJ, Tolias AS. A neural basis of probabilistic computation in visual cortex. Nat Neurosci. 2020. January;23(1):122–9. 10.1038/s41593-019-0554-5 [DOI] [PubMed] [Google Scholar]
15.Casella G, Berger RL. Statistical inference. Cengage Learning; 2021. January 26. [Google Scholar]
16.Wei XX, Stocker AA. Lawful relation between perceptual bias and discriminability. Proc Natl Acad Sci. 2017. September 19;114(38):10244–9. 10.1073/pnas.1619153114 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Campbell FW, Kulikowski JJ, Levinson J. The effect of orientation on the visual resolution of gratings. J Physiol. 1966. November 1;187(2):427–36. 10.1113/jphysiol.1966.sp008100 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Appelle S. Perception and discrimination as a function of stimulus orientation: the" oblique effect" in man and animals. Psychol Bull. 1972. October;78(4):266. 10.1037/h0033117 [DOI] [PubMed] [Google Scholar]
19.Westheimer G, Beard BL. Orientation dependency for foveal line stimuli: detection and intensity discrimination, resolution, orientation discrimination and vernier acuity. Vision Res. 1998. April 1;38(8):1097–103. 10.1016/s0042-6989(97)00248-4 [DOI] [PubMed] [Google Scholar]
20.Dakin SC, Mareschal I, Bex PJ. An oblique effect for local motion: Psychophysics and natural movie statistics. J Vis. 2005. November 1;5(10):9. 10.1167/5.10.9 [DOI] [PubMed] [Google Scholar]
21.Girshick AR, Landy MS, Simoncelli EP. Cardinal rules: visual orientation perception reflects knowledge of environmental statistics. Nat Neurosci. 2011. July;14(7):926–32. 10.1038/nn.2831 [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Coppola DM, Purves HR, McCoy AN, Purves D. The distribution of oriented contours in the real world. Proc Natl Acad Sci. 1998. March 31;95(7):4002–6. 10.1073/pnas.95.7.4002 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Wei XX, Stocker AA. A Bayesian observer model constrained by efficient coding can explain’anti-Bayesian’percepts. Nat Neurosci. 2015. October;18(10):1509. 10.1038/nn.4105 [DOI] [PubMed] [Google Scholar]
24.Vogels R, Orban GA. The effect of practice on the oblique effect in line orientation judgments. Vision Res. 1985. January 1;25(11):1679–87. 10.1016/0042-6989(85)90140-3 [DOI] [PubMed] [Google Scholar]
25.Shiu LP, Pashler H. Improvement in line orientation discrimination is retinally local but dependent on cognitive set. Percept Psychophys. 1992. September;52(5):582–8. 10.3758/bf03206720 [DOI] [PubMed] [Google Scholar]
26.Schwartz O, Hsu A, Dayan P. Space and time in visual context. Nat Rev Neurosci. 2007. July;8(7):522–35. 10.1038/nrn2155 [DOI] [PubMed] [Google Scholar]
27.Milne E. Increased intra-participant variability in children with autistic spectrum disorders: evidence from single-trial analysis of evoked EEG. Front Psychol. 2011. March 30;2:51. 10.3389/fpsyg.2011.00051 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Dinstein I, Heeger DJ, Lorenzi L, Minshew NJ, Malach R, Behrmann M. Unreliable evoked responses in autism. Neuron. 2012. September 20;75(6):981–91. 10.1016/j.neuron.2012.07.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Haigh SM, Heeger DJ, Dinstein I, Minshew N, Behrmann M. Cortical variability in the sensory-evoked response in autism. J Autism Dev Disord. 2015. May 1;45(5):1176–90. 10.1007/s10803-014-2276-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Zaidel A, Goin-Kochel RP, Angelaki DE. Self-motion perception in autism is compromised by visual noise but integrated optimally across multiple senses. Proc Natl Acad Sci. 2015. May 19;112(20):6461–6. 10.1073/pnas.1506582112 [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Noel JP, Lakshminarasimhan KJ, Park H, Angelaki DE. Increased variability but intact integration during visual navigation in Autism Spectrum Disorder. Proc Natl Acad Sci. 2020. May 19;117(20):11158–66. 10.1073/pnas.2000216117 [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Seriès P, Stocker AA, Simoncelli EP. Is the homunculus “aware” of sensory adaptation?. Neural Comput. 2009. December;21(12):3271–304. 10.1162/neco.2009.09-08-869 [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Barlow HB. Possible principles underlying the transformation of sensory messages. Sensory Communication. 1961. September;1(01). [Google Scholar]
34.Młynarski WF, Hermundstad AM. Adaptive coding for dynamic sensory inference. Elife. 2018. July 10;7:e32055. 10.7554/eLife.32055 [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Młynarski W, Hermundstad AM. Adaptability and efficiency in neural coding. bioRxiv. 2019. January 1:669200. [Google Scholar]
36.Simmons DR, Robertson AE, McKay LS, Toal E, McAleer P, Pollick FE. Vision in autism spectrum disorders. Vision Res 2009. November 10;49(22):2705–39. 10.1016/j.visres.2009.08.005 [DOI] [PubMed] [Google Scholar]
37.Park WJ, Schauder KB, Zhang R, Bennetto L, Tadin D. High internal noise and poor external noise filtering characterize perception in autism spectrum disorder. Sci Rep. 2017. December 14;7(1):1–2. 10.1038/s41598-016-0028-x [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Lu ZL, Dosher BA. Characterizing observers using external noise and observer models: assessing internal representations with external noise. Psychol Rev. 2008. January;115(1):44. 10.1037/0033-295X.115.1.44 [DOI] [PubMed] [Google Scholar]
39.Lu ZL, Dosher BA. External noise distinguishes attention mechanisms. Vision Res. 1998. May 1;38(9):1183–98. 10.1016/s0042-6989(97)00273-3 [DOI] [PubMed] [Google Scholar]
40.Goldstone RL. Perceptual learning. Annu Rev Psychol. 1998. February;49(1):585–612. 10.1146/annurev.psych.49.1.585 [DOI] [PubMed] [Google Scholar]
41.Maniglia M, Seitz AR. Towards a whole brain model of Perceptual Learning. Curr Opin Behav Sci. 2018. April 1;20:47–55. 10.1016/j.cobeha.2017.10.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Harris H, Israeli D, Minshew N, Bonneh Y, Heeger DJ, Behrmann M, et al. Perceptual learning in autism: over-specificity and possible remedies. Nat Neurosci. 2015;18 (11):1574–6. 10.1038/nn.4129 [DOI] [PubMed] [Google Scholar]
43.Schoups A, Vogels R, Qian N, Orban G. Practising orientation identification improves orientation coding in V1 neurons. Nature. 2001. August;412(6846):549–53. 10.1038/35087601 [DOI] [PubMed] [Google Scholar]
44.Ganguli D, Simoncelli EP. Efficient sensory encoding and Bayesian inference with heterogeneous neural populations. Neural Comput. 2014. October;26(10):2103–34. 10.1162/NECO_a_00638 [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Clifford CW, Webster MA, Stanley GB, Stocker AA, Kohn A, Sharpee TO, et al. Visual adaptation: Neural, psychological and computational aspects. Vision Res. 2007. November 1;47(25):3125–31. 10.1016/j.visres.2007.08.023 [DOI] [PubMed] [Google Scholar]
46.Cohen MR, Maunsell JH. Attention improves performance primarily by reducing interneuronal correlations. Nat Neurosci. 2009. December;12(12):1594. 10.1038/nn.2439 [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Coen-Cagli R, Solomon SS. Relating divisive normalization to neuronal response variability. J Neurosci. 2019. September 11;39(37):7344–56. 10.1523/JNEUROSCI.0126-19.2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Carandini M, Heeger DJ. Normalization as a canonical neural computation. Nat Rev Neurosci. 2012. January;13(1):51–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Rosenberg A, Patterson JS, Angelaki DE. A computational perspective on autism. Proc Natl Acad Sci. 2015. July 28;112(30):9158–65. 10.1073/pnas.1510583112 [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Li B, Peterson MR, Freeman RD. Oblique effect: a neural basis in the visual cortex. J Neurophysiol. 2003. July;90(1):204–17. 10.1152/jn.00954.2002 [DOI] [PubMed] [Google Scholar]
51.Gowen E, Hamilton A. Motor abilities in autism: a review using a computational context. J Autism Dev Disord. 2013. February 1;43(2):323–44. 10.1007/s10803-012-1574-0 [DOI] [PubMed] [Google Scholar]
52.Luu L, Stocker AA. Post-decision biases reveal a self-consistency principle in perceptual inference. Elife. 2018. May 22;7:e33334. 10.7554/eLife.33334 [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Stocker AA, Simoncelli EP. Noise characteristics and prior expectations in human visual speed perception. Nat Neurosci. 2006. April;9(4):578–85. 10.1038/nn1669 [DOI] [PubMed] [Google Scholar]
54.Karvelis P, Seitz AR, Lawrie SM, Seriès P. Autistic traits, but not schizotypy, predict increased weighting of sensory information in Bayesian visual integration. Elife. 2018. May 14;7:e34115. 10.7554/eLife.34115 [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Coskun MA, Varghese L, Reddoch S, Castillo EM, Pearson DA, Loveland KA, Papanicolaou AC, Sheth BR. Increased response variability in autistic brains?. Neuroreport. 2009. November 25;20(17):1543–8. 10.1097/WNR.0b013e32833246b5 [DOI] [PubMed] [Google Scholar]
56.Manning C, Tibber MS, Dakin SC. Visual integration of direction and orientation information in autistic children. Autism & Developmental Language Impairments. 2017. February;2:2396941517694626. [Google Scholar]
57.Manning C, Tibber MS, Charman T, Dakin SC, Pellicano E. Enhanced integration of motion information in children with autism. J Neurosci. 2015. May 6;35(18):6979–86. 10.1523/JNEUROSCI.4645-14.2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Butler JS, Molholm S, Andrade GN, Foxe JJ. An examination of the neural unreliability thesis of autism. Cereb Cortex. 2017. January 1;27(1):185–200. 10.1093/cercor/bhw375 [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Baron-Cohen S, Wheelwright S, Skinner R, Martin J, Clubley E. The autism-spectrum quotient (AQ): Evidence from asperger syndrome/high-functioning autism, malesand females, scientists and mathematicians. J Autism Dev Disord. 2001. February;31(1):5–17. 10.1023/a:1005653411471 [DOI] [PubMed] [Google Scholar]
60.Rutter M, Bailey A, Lord C. The social communication questionnaire: Manual. Western Psychological Services. 2003. [Google Scholar]
61.Efron B, Tibshirani RJ. An introduction to the bootstrap. CRC press; 1994. May 15. [Google Scholar]

PLoS Biol. doi: 10.1371/journal.pbio.3001215.r001

Decision Letter 0

Lucas Smith

15 Sep 2020

Dear Dr Noel,

Thank you for submitting your manuscript entitled "Aberrant Sensory Encoding in Patients with Autism" for consideration as a Research Article by PLOS Biology.

Your manuscript has now been evaluated by the PLOS Biology editorial staff as well as by an academic editor with relevant expertise and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by Sep 17 2020 11:59PM.

During resubmission, you will be invited to opt-in to posting your pre-review manuscript as a bioRxiv preprint. Visit http://journals.plos.org/plosbiology/s/preprints for full details. If you consent to posting your current manuscript as a preprint, please upload a single Preprint PDF when you re-submit.

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Given the disruptions resulting from the ongoing COVID-19 pandemic, please expect delays in the editorial process. We apologise in advance for any inconvenience caused and will do our best to minimize impact as far as possible.

Feel free to email us at plosbiology@plos.org if you have any queries relating to your submission.

Kind regards,

Lucas Smith, Ph.D.,

Associate Editor

PLOS Biology

PLoS Biol. doi: 10.1371/journal.pbio.3001215.r002

Decision Letter 1

Lucas Smith

12 Nov 2020

Dear Dr Noel,

Thank you very much for submitting your manuscript "Aberrant Sensory Encoding in Patients with Autism" for consideration as a Research Article at PLOS Biology. Your manuscript has been evaluated by the PLOS Biology editors, an Academic Editor with relevant expertise, and by several independent reviewers.

The reviews of your manuscript are appended below. As you will see from their comments, all of the reviewers felt that the topic of this study was interesting, however the reviewers have raised potential issues with the experimental paradigm, and feel that in places the conclusions of the study are not fully supported by the data.

In light of the reviews, we will not be able to accept the current version of the manuscript, but we would welcome re-submission of a much-revised version that takes into account the reviewers' comments and addresses the specific points raised by the reviewers. Please pay special attention to the concerns raised by Reviewer 1, which should be thoroughly and satisfactorily addressed.

We expect to receive your revised manuscript within 3 months.

Please email us (plosbiology@plos.org) if you have any questions or concerns, or would like to request an extension. At this stage, your manuscript remains formally under active consideration at our journal; please notify us by email if you do not intend to submit a revision so that we may end consideration of the manuscript at PLOS Biology.

**IMPORTANT - SUBMITTING YOUR REVISION**

Your revisions should address the specific points made by each reviewer. Please submit the following files along with your revised manuscript:

1. A 'Response to Reviewers' file - this should detail your responses to the editorial requests, present a point-by-point response to all of the reviewers' comments, and indicate the changes made to the manuscript.

*NOTE: In your point by point response to the reviewers, please provide the full context of each review. Do not selectively quote paragraphs or sentences to reply to. The entire set of reviewer comments should be present in full and each specific point should be responded to individually, point by point.

You should also cite any additional relevant literature that has been published since the original submission and mention any additional citations in your response.

2. In addition to a clean copy of the manuscript, please also upload a 'track-changes' version of your manuscript that specifies the edits made. This should be uploaded as a "Related" file type.

*Re-submission Checklist*

When you are ready to resubmit your revised manuscript, please refer to this re-submission checklist: https://plos.io/Biology_Checklist

To submit a revised version of your manuscript, please go to https://www.editorialmanager.com/pbiology/ and log in as an Author. Click the link labelled 'Submissions Needing Revision' where you will find your submission record.

Please make sure to read the following important policies and guidelines while preparing your revision:

*Published Peer Review*

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. Please see here for more details:

https://blogs.plos.org/plos/2019/05/plos-journals-now-open-for-published-peer-review/

*PLOS Data Policy*

Please note that as a condition of publication PLOS' data policy (http://journals.plos.org/plosbiology/s/data-availability) requires that you make available all data used to draw the conclusions arrived at in your manuscript. If you have not already done so, you must include any data used in your manuscript either in appropriate repositories, within the body of the manuscript, or as supporting information (N.B. this includes any numerical values that were used to generate graphs, histograms etc.). For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5

*Blot and Gel Data Policy*

We require the original, uncropped and minimally adjusted images supporting all blot and gel results reported in an article's figures or Supporting Information files. We will require these files before a manuscript can be accepted so please prepare them now, if you have not already uploaded them. Please carefully read our guidelines for how to prepare and upload this data: https://journals.plos.org/plosbiology/s/figures#loc-blot-and-gel-reporting-requirements

*Protocols deposition*

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosbiology/s/submission-guidelines#loc-materials-and-methods

Thank you again for your submission to our journal. We hope that our editorial process has been constructive thus far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Lucas Smith, Ph.D.,

Associate Editor,

lsmith@plos.org,

PLOS Biology

*****************************************************

REVIEWS:

Reviewer's Responses to Questions

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Albert Powers

Reviewer #3: Yes: Duje Tadin

Reviewer #1: The authors use the oblique effect to examine sensory encoding in neurotypical and autistic individuals. They report that after undergoing trials with feedback, controls, but not autistic participants change their sensory representations and 'increase encoding capacity'. This, they conclude, suggests deficits in sensory encoding in autism and compromised ability to adapt to novel input statistics. I find the notion of autism being associated with inflexible priors quite appealing, given its potential explanatory scope for many facets of the phenotype. However, I am not persuaded that the empirical evidence presented here provides unequivocal support for this proposal. My concerns are related to the experimental paradigm used and also the interpretation of some of the findings.

The oblique effect strikes me as a less than ideal phenomenon to use for probing dynamic changes in priors. Although the neural basis of this effect is not definitively established, studies that have examined responses from large collections of neurons in visual cortex suggest that anisotropies in cell preferences as early as V1 may largely be responsible for the phenomenon (e.g. Baowang Li, Matthew R. Peterson, and Ralph D. Freeman (2003). Oblique Effect: A Neural Basis in the Visual Cortex. J Neurophysiol 90: 204-217). Response properties of neurons in early sensory cortex, that have been shaped by long evolutionary history, are not readily modifiable by short experiences. Hence, the oblique effect may not be the ideal vehicle to assess the impact of short-term changes in image statistics on sensory encoding.

The task that participants are asked to perform relies not just upon encoding, but also short-term memory (to remember the seen orientation until a response is made) and the ability to avoid 'contamination' of the remembered orientation by the perceived orientation of the response Gabor itself. In this context, can the differences in performance between controls and ASD participants be unambiguously ascribed to encoding differences alone? It is possible that ASD participants are more susceptible to contamination between response probe orientation and remembered orientation, and hence do not show the kinds of gains in performance that controls do across the experimental blocks. (I might have missed this in the manuscript, but it is important to state how long it takes subjects in the two groups to respond, i.e. reorient the probe Gabor from its random starting orientation to the final one. How monotonic is this process? Do subjects exhibit overshoots and corrections? Do the different groups differ in this? Might ASD participants be more or less impulsive in their responses relative to controls and might this impact their response bias? All of this is to make the point that the process of responding may be somewhat protracted and involve inadvertent modifications of the remembered target orientation.)

I am also skeptical of the suggestion that controls' priors are changed by exposure to stimuli with a uniform distribution of orientation. If that were to be the case, exposure alone should have sufficed; feedback should not have been necessary. Since anisotropies in orientation encoding are believed to arise from mere exposure to a world with increased representation of cardinal directions, why should the modification of these priors require feedback?

On a minor note, the classic oblique effect is associated with high orientation discrimination accuracy for horizontal and vertical orientations ('cardinal' directions) and least for obliques. In the data reported here, cardinal as well as oblique directions elicit high accuracies. What might account for this discrepancy? Might it be an outcome of the memory component of the task used here, with the four orientations being recalled symbolically ('horizontal', 'vertical', 'left diagonal', 'right diagonal')?

Overall, while I am sympathetic to the conceptual point the manuscript is attempting to make, I am not entirely convinced that the empirical data are adequately clear in the interpretations they allow.

Reviewer #2: In this ambitious paper, the authors attempt to take advantage of the so-called "oblique effect" to estimate aberrant sensory encoding in ASD. This is a really interesting and clever approach, but I have several concerns that significantly impact my enthusiasm for the manuscript in its current form. These are enumerated below.

1. Authors make the case that encoding and decoding can be isolated behaviorally. However, the fullness of pathophysiology may come from the pairing between aberrant sensory encoding and some decoding scheme.

2. Relatedly, others have demonstrated high-precision likelihoods in ASD. Why would this arise in the context of low-fidelity sensory encoding? Wouldn't one expect the opposite? Or is it the authors' contention that these two deficits are unrelated? This should be explained more fully in the Discussion.

3. Bias and variance (Fig 1), observed behaviorally, are taken to be direct read-outs of the percept resulting from decoding. But of course this cannot be—a response is generated somehow and a response model must be specified if behavioral responses are used to estimate the quantities included in the model. What is this response model? If there is none,why not?

4. The authors would do well to motivate the use of feedback blocks during the introduction. Feedback comes somewhat as a surprise when introduced in the Results section and

the reader is unclear why this manipulation was introduced in the context of the overall study.

5. Related to 3, and especially in the context of feedback-based learning, it is not surprising that he authors how an improvement in FI while showing a decrease in bias. There are no other variables capable of explaining that difference, given the authors' model. This doesn't mean that the results are not due to simple response bias or any other post-encoding step of processing, but rather that the model, in its incompleteness, is incapable of accounting for any other reason for the behavioral results. The authors provide reasons why behavioral noise and decoding processes are unlikely to account for the results observed. But the only way to make this truly convincing is to test it formally by including a Bayesian decoding and response model with relevant parameters included.

Reviewer 3: This study investigates sensory encoding in autism (ASD). Over the past decade, there has been an increase in work looking at sensory abnormalities in autism, but we still lack a good understanding of how sensory, namely visual, processing is atypical in autism. This study focuses on visual encoding, a critical early step in visual processing. This question is addressed in the context of visual orientation discrimination and combines convincing behavioral data (I’d say remarkably convincing given what is typically seen in ASD work) and sophisticated computational modeling. The conclusion is that sensory encoding is severely impaired in ASD.

I find this to be an interesting and a noteworthy study. The behavioral data are, to repeat, convincing and certainly indicate atypical orientation processing in ASD. The modeling is well done, and while I’m not ready to fully “buy” the ultimate conclusions (details below), even if those conclusions turn out to be not entirely correct, this work will still be an important addition to our understanding of how sensory processing could be impaired in ASD.

1A. There are two key behavioral findings: (1) the overall difference in performance in the first block and (2) poor learning/adaptation of the ASD group over blocks. Both findings need to be better integrated with the existing literature, both to better contextualize the present results and to consider alternative explanations. The authors do link the first result to other studies showing increased variability in ASD, but more is needed there. For example, a recent study that examined variability of visual orientation processing is not mentioned. Also, it is worth mentioning work that questions the “noise hypothesis” in ASD. Both references are below.

Park, W. J., Schauder, K. B., Zhang, R., Bennetto, L., & Tadin, D. (2017). High internal noise and poor external noise filtering characterize perception in autism spectrum disorder. Scientific reports, 7(1), 1-12.

Butler, J. S., Molholm, S., Andrade, G. N., & Foxe, J. J. (2017). An examination of the neural unreliability thesis of autism. Cerebral cortex, 27(1), 185-200.

1B. In addition, the change in performance over blocks is discussed as learning, but no work on perceptual learning in ASD is mentioned. For example, the paper listed below showed that perceptual learning in ASD is atypical. This raises the question of whether the present results could be a result of a boarder deficit in learning?

Harris, H., Israeli, D., Minshew, N., Bonneh, Y., Heeger, D. J., Behrmann, M., & Sagi, D. (2015). Perceptual learning in autism: over-specificity and possible remedies. Nature neuroscience, 18(11), 1574-1576.

2. The first block does not use feedback, while the 2nd and the 3rd blocks do. That decision needs to be explained and justified, but it is not. Why was adding feedback after the first block important? After all, feedback is not necessary for us to learn stimulus statistics (though it likely speeds up the process).

My worry is that this leaves open a possibility that how the two groups responded to this change in feedback

could be, in part, what drives the difference. For example, what if individuals with ASD ignored feedback? (btw, this is a hypothetical, I don’t have a reason to believe this). What I’m looking for here is a better justification of why the feedback was used in this specific way and are there good reasons to believe that the results would generalize to other ways of deploying feedback (or no feedback at all). That is, I’m not asking for new data (certainly not during COVID times).

3. The AQ results (Fig 5) are entirely un-informative. The reported correlations seem to be mostly driven by the overall group differences, with little evidence of within-group correlation effects. For these conclusions to stand, what is needed is evidence that **within** the ASD group, these results are corelates with the symptom severity.

Moreover, AQ is a somewhat outdated measure that relies a lot on imperfect stereotypes of what autism is, and it is not used in the diagnosis process (rather it is best used as a measure of autism-like behaviors and thinking among the general population). However, it does look like ADOS was collected.

Minor points:

-- I’d take out “patients” from the title. Many individuals with autism do not consider themselves to be patients.

-- In discussion (page 9), it is stated “This speculation will have to be confirmed in physiological studies.” It is not clear how. Are we talking about EEG and fMRI or animal models?

-- Materials and Procedures: stimulus sizes need to be given in visual degrees and not centimeters.

Attachment

Submitted filename: Review.pdf

Click here for additional data file.^{(111.8KB, pdf)}

PLoS Biol. 2021 May 12;19(5):e3001215. doi: 10.1371/journal.pbio.3001215.r003

Author response to Decision Letter 1

10 Feb 2021

Attachment

Submitted filename: PlosBio_Letter_to_reviewers.docx

Click here for additional data file.^{(1.5MB, docx)}

PLoS Biol. doi: 10.1371/journal.pbio.3001215.r004

Decision Letter 2

Lucas Smith

23 Mar 2021

Dear Dr Noel,

Thank you for submitting your revised Research Article entitled "Aberrant Visual Orientation Encoding in Individuals with Autism" for publication in PLOS Biology. I have now obtained advice from the original reviewers and have discussed their comments with the Academic Editor.

The reviewers are all satisfied with the revised manuscript and think their previous concerns have been satisfactorily addressed. Therefore we will probably accept this manuscript for publication, provided you address the remaining editorial, data, and other policy-related requests included here and in more detail below my signature.

1) Thank you for adapting your title in response to reviewer 3's comment that "Many individuals with autism do not consider themselves to be patients". Along these lines, we think the rest of the manuscript should be edited to reflect this. We think language calling individuals with ASD 'patients' should be changed throughout the manuscript and abstract. Similarly, we think that instead of calling the control group 'healthy' you might use 'neurotypical'.

2) We would also like to suggest an edit to the title, which we think might help make it a bit more direct and clear. If you agree, we think the title could be changed to "People with autism spectrum disorder have altered visual encoding capacity".

3) Ethics request: Please indicate whether your protocol, approved by the Baylor College and Medicine IRB adhered to the Declaration of Helsinki or other national or international ethical guidelines.

4) Ethics Request: Please include the ID number of the protocol approved by the Baylor College and Medicine IRB.

5) Data Request: Please provide, as a supplementary file, the summary statistics used to generate each graph presented in your study. Please make sure to provide a legend for this file, and to refer to it in your data availability statement and figure legends (including supplementary figure legends). For example, to each figure legend, you could add a statement saying “The data underlying this figure can be found in the supplementary file S1_data”.

As you address these items, please take this last chance to review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the cover letter that accompanies your revised manuscript.

We expect to receive your revised manuscript within two weeks.

To submit your revision, please go to https://www.editorialmanager.com/pbiology/ and log in as an Author. Click the link labelled 'Submissions Needing Revision' to find your submission record. Your revised submission must include the following:

- a cover letter that should detail your responses to any editorial requests, if applicable, and whether changes have been made to the reference list

- a Response to Reviewers file that provides a detailed response to the reviewers' comments (if applicable)

- a track-changes file indicating any changes that you have made to the manuscript.

NOTE: If Supporting Information files are included with your article, note that these are not copyedited and will be published as they are submitted. Please ensure that these files are legible and of high quality (at least 300 dpi) in an easily accessible file format. For this reason, please be aware that any references listed in an SI file will not be indexed. For more information, see our Supporting Information guidelines:

https://journals.plos.org/plosbiology/s/supporting-information

*Published Peer Review History*

Please note that you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. Please see here for more details:

https://blogs.plos.org/plos/2019/05/plos-journals-now-open-for-published-peer-review/

*Early Version*

Please note that an uncorrected proof of your manuscript will be published online ahead of the final version, unless you opted out when submitting your manuscript. If, for any reason, you do not want an earlier version of your manuscript published online, uncheck the box. Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us as soon as possible if you or your institution is planning to press release the article.

*Protocols deposition*

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please do not hesitate to contact me should you have any questions.

Sincerely,

Lucas Smith, Ph.D.,

Associate Editor,

lsmith@plos.org,

PLOS Biology

------------------------------------------------------------------------

ETHICS STATEMENT:

-- Please indicate whether your protocol, approved by the Baylor College and Medicine IRB adhered to the Declaration of Helsinki or other national or international ethical guidelines

-- Please include the ID number of the protocol approved by the Baylor College and Medicine IRB.

------------------------------------------------------------------------

DATA POLICY:

You may be aware of the PLOS Data Policy, which requires that all data be made available without restriction: http://journals.plos.org/plosbiology/s/data-availability. For more information, please also see this editorial: http://dx.doi.org/10.1371/journal.pbio.1001797

Note that we do not require all raw data. Rather, we ask that all individual quantitative observations that underlie the data summarized in the figures and results of your paper be made available in one of the following forms:

1) Supplementary files (e.g., excel). Please ensure that all data files are uploaded as 'Supporting Information' and are invariably referred to (in the manuscript, figure legends, and the Description field when uploading your files) using the following format verbatim: S1 Data, S2 Data, etc. Multiple panels of a single or even several figures can be included as multiple sheets in one excel file that is saved using exactly the following convention: S1_Data.xlsx (using an underscore).

2) Deposition in a publicly available repository. Please also provide the accession code or a reviewer link so that we may view your data before publication.

Regardless of the method selected, please ensure that you provide the individual numerical values that underlie the summary data displayed in the following figure panels as they are essential for readers to assess your analysis and to reproduce it:

Fig 1d-f; Fig 2B-C; Fig 3A-F; Fig 4 A-B,D-E; Fig 5A-B; Fig S1; Fig S2; Fig S3; Fig S4; Fig S5; Fig S6; Fig S7B; Fig S8A-D

NOTE: the numerical data provided should include all replicates AND the way in which the plotted mean and errors were derived (it should not present only the mean/average values).

Please also ensure that figure legends in your manuscript include information on where the underlying data can be found, and ensure your supplemental data file/s has a legend.

Please ensure that your Data Statement in the submission system accurately describes where your data can be found.

------------------------------------------------------------------------

Reviewer remarks:

Reviewer #1: The authors have thoughtfully responded to the concerns I had raised in my earlier review. I believe the manuscript is now suitable for publication.

Reviewer #2: The authors have done an impressive job of responding to my comments comprehensively. I have no further comments.

Reviewer #3 (Duje Tadin): I appreciate the authors' careful work on this revision. It addresses the concerns I had about this manuscript, namely concerns about links with other literature on sensory noise and on learning. I also appreciate the detailed consideration and additional analyses on the role of feedback, and how the authors responded to concerns about the AQ analysis.

PLoS Biol. 2021 May 12;19(5):e3001215. doi: 10.1371/journal.pbio.3001215.r005

Author response to Decision Letter 2

29 Mar 2021

Attachment

Submitted filename: PlosBio_Letter_to_reviewers.docx

Click here for additional data file.^{(23.3KB, docx)}

PLoS Biol. doi: 10.1371/journal.pbio.3001215.r006

Decision Letter 3

Lucas Smith

1 Apr 2021

Dear Dr Noel,

Thank you very much for addressing our last editorial requests in a revised manuscript. On behalf of my colleagues and the Academic Editor, Adam Kohn, I am pleased to say that we can now, in principle offer to publish your Research Article "Individuals with Autism Spectrum Disorder Have Altered Visual Encoding Capacity" in PLOS Biology, provided you address any remaining formatting and reporting issues. These will be detailed in an email that will follow this letter and that you will usually receive within 2-3 business days, during which time no action is required from you. Please note that we will not be able to formally accept your manuscript and schedule it for publication until you have made the required changes.

When addressing these last formatting requests, we also ask that adjust your data availability statement to reference the recently added supplementary files. For example, you could say “The data underlying each figure can be found in supplementary files. Additional data and code for our analysis can be found at: https://github.com/cpc-lab-stocker/ASD_Encoding_2020”

Please take a minute to log into Editorial Manager at http://www.editorialmanager.com/pbiology/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with biologypress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

Thank you again for supporting Open Access publishing. We look forward to publishing your paper in PLOS Biology.

Sincerely,

Lucas Smith, Ph.D.

Associate Editor

PLOS Biology

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Text. Motivation and analytical derivation for using the Cramer–Rao lower bound to measure Fisher information of the encoding process.

(DOCX)

Click here for additional data file.^{(40.4KB, docx)}

S1 Fig

(EPS)

Click here for additional data file.^{(17.7MB, eps)}

S2 Fig

(EPS)

Click here for additional data file.^{(12.1MB, eps)}

S3 Fig

(EPS)

Click here for additional data file.^{(2MB, eps)}

S4 Fig

(EPS)

Click here for additional data file.^{(2MB, eps)}

S5 Fig. Histogram of RT (i.e., time it took for participants to rotate the response Gabor patch to match the target orientation).

(EPS)

Click here for additional data file.^{(2.5MB, eps)}

S6 Fig

(EPS)

Click here for additional data file.^{(3.6MB, eps)}

S7 Fig

(EPS)

Click here for additional data file.^{(3MB, eps)}

(EPS)

Click here for additional data file.^{(2.8MB, eps)}

S1 Code. Raw data and MATLAB code for our analysis and figures.

Routine to reproduce Figs 1D–1F, 2B, 2C, 3, 4 and 5 and S1–S6, panel B in S7, and S8 Figs.

(ZIP)

Click here for additional data file.^{(1.1MB, zip)}

S1 Data. Numerical values that make up Fig 1D–1F.

(ZIP)

Click here for additional data file.^{(5.8MB, zip)}

S2 Data. Numerical values that make up Fig 2B and 2C.

(ZIP)

Click here for additional data file.^{(82.4KB, zip)}

S3 Data. Numerical values that make up Fig 3.

(ZIP)

Click here for additional data file.^{(19MB, zip)}

S4 Data. Numerical values that make up Fig 4.

(ZIP)

Click here for additional data file.^{(13.3MB, zip)}

S5 Data. Numerical values that make up Fig 5 and S3 and S4 Figs.

(ZIP)

Click here for additional data file.^{(16.7KB, zip)}

S6 Data. Numerical values that make up S1, S2, S5, S6, panel B in S7, and S8 Figs.

(ZIP)

Click here for additional data file.^{(6.6MB, zip)}

Attachment

Submitted filename: Review.pdf

Click here for additional data file.^{(111.8KB, pdf)}

Attachment

Submitted filename: PlosBio_Letter_to_reviewers.docx

Click here for additional data file.^{(1.5MB, docx)}

Attachment

Submitted filename: PlosBio_Letter_to_reviewers.docx

Click here for additional data file.^{(23.3KB, docx)}

Data Availability Statement

The data underlying each figure can be found in supplementary files. Additional data and code for our analysis can be found at: https://github.com/cpc-lab-stocker/ASD_Encoding_2020.

[pbio.3001215.ref001] 1.Xu G, Strathearn L, Liu B, O’Brien M, Kopelman TG, Zhu J, et al. Prevalence and Treatment Patterns of Autism Spectrum Disorder in the United States, 2016. JAMA Pediatr. 2019. February 1;173(2):153–9. 10.1001/jamapediatrics.2018.4208 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref002] 2.Pellicano E, Burr D. When the world becomes ’too real’: a Bayesian explanation of autistic perception. Trends Cogn Sci. 2012. October;16(10):504–10. 10.1016/j.tics.2012.08.009 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref003] 3.Brock J. Alternative Bayesian accounts of autistic perception: comment on Pellicano and Burr. Trends Cogn Sci. 2012. December;16(12):573–4; author reply 574–5. 10.1016/j.tics.2012.10.005 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref004] 4.Van de Cruys S, Evers K, Van der Hallen R, Van Eylen L, Boets B, de-Wit L, et al. Precise minds in uncertain worlds: predictive coding in autism. Psychol Rev. 2014. Oct;121(4):649–75. 10.1037/a0037665 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref005] 5.Lawson RP, Rees G, Friston KJ. An aberrant precision account of autism. Front Hum Neurosci. 2014;8:302. 10.3389/fnhum.2014.00302 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref006] 6.Powell G, Meredith Z, McMillin R, Freeman TC. Bayesian models of individual differences: Combining autistic traits and sensory thresholds to predict motion perception. Psychol Sci. 2016. December;27(12):1562–72. 10.1177/0956797616665351 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref007] 7.Karaminis T, Cicchini GM, Neil L, Cappagli G, Aagten-Murphy D, Burr D, et al. Central tendency effects in time interval reproduction in autism. Sci Rep. 2016. June 28;6(1):1–3. 10.1038/s41598-016-0001-8 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref008] 8.Haker H, Schneebeli M, Stephan KE. Can Bayesian theories of autism spectrum disorder help improve clinical practice?. Front Psych. 2016. June 17;7:107. 10.3389/fpsyt.2016.00107 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref009] 9.Palmer CJ, Lawson RP, Hohwy J. Bayesian approaches to autism: Towards volatility, action, and behavior. Psychol Bull. 2017. May;143(5):521. 10.1037/bul0000097 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref010] 10.Lawson RP, Mathys C, Rees G. Adults with autism overestimate the volatility of the sensory environment. Nat Neurosci. 2017. September;20(9):1293. 10.1038/nn.4615 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref011] 11.Lieder I, Adam V, Frenkel O, Jaffe-Dax S, Sahani M, Ahissar M. Perceptual bias reveals slow-updating in autism and fast-forgetting in dyslexia. Nat Neurosci. 2019. February;22(2):256–64. 10.1038/s41593-018-0308-9 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref012] 12.Ernst MO, Banks MS. Humans integrate visual and haptic information in a statistically optimal fashion. Nature. 2002. January;415(6870):429–33. 10.1038/415429a [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref013] 13.Hung CP, Kreiman G, Poggio T, DiCarlo JJ. Fast readout of object identity from macaque inferior temporal cortex. Science. 2005. November 4;310(5749):863–6. 10.1126/science.1117593 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref014] 14.Walker EY, Cotton RJ, Ma WJ, Tolias AS. A neural basis of probabilistic computation in visual cortex. Nat Neurosci. 2020. January;23(1):122–9. 10.1038/s41593-019-0554-5 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref015] 15.Casella G, Berger RL. Statistical inference. Cengage Learning; 2021. January 26. [Google Scholar]

[pbio.3001215.ref016] 16.Wei XX, Stocker AA. Lawful relation between perceptual bias and discriminability. Proc Natl Acad Sci. 2017. September 19;114(38):10244–9. 10.1073/pnas.1619153114 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref017] 17.Campbell FW, Kulikowski JJ, Levinson J. The effect of orientation on the visual resolution of gratings. J Physiol. 1966. November 1;187(2):427–36. 10.1113/jphysiol.1966.sp008100 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref018] 18.Appelle S. Perception and discrimination as a function of stimulus orientation: the" oblique effect" in man and animals. Psychol Bull. 1972. October;78(4):266. 10.1037/h0033117 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref019] 19.Westheimer G, Beard BL. Orientation dependency for foveal line stimuli: detection and intensity discrimination, resolution, orientation discrimination and vernier acuity. Vision Res. 1998. April 1;38(8):1097–103. 10.1016/s0042-6989(97)00248-4 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref020] 20.Dakin SC, Mareschal I, Bex PJ. An oblique effect for local motion: Psychophysics and natural movie statistics. J Vis. 2005. November 1;5(10):9. 10.1167/5.10.9 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref021] 21.Girshick AR, Landy MS, Simoncelli EP. Cardinal rules: visual orientation perception reflects knowledge of environmental statistics. Nat Neurosci. 2011. July;14(7):926–32. 10.1038/nn.2831 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref022] 22.Coppola DM, Purves HR, McCoy AN, Purves D. The distribution of oriented contours in the real world. Proc Natl Acad Sci. 1998. March 31;95(7):4002–6. 10.1073/pnas.95.7.4002 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref023] 23.Wei XX, Stocker AA. A Bayesian observer model constrained by efficient coding can explain’anti-Bayesian’percepts. Nat Neurosci. 2015. October;18(10):1509. 10.1038/nn.4105 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref024] 24.Vogels R, Orban GA. The effect of practice on the oblique effect in line orientation judgments. Vision Res. 1985. January 1;25(11):1679–87. 10.1016/0042-6989(85)90140-3 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref025] 25.Shiu LP, Pashler H. Improvement in line orientation discrimination is retinally local but dependent on cognitive set. Percept Psychophys. 1992. September;52(5):582–8. 10.3758/bf03206720 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref026] 26.Schwartz O, Hsu A, Dayan P. Space and time in visual context. Nat Rev Neurosci. 2007. July;8(7):522–35. 10.1038/nrn2155 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref027] 27.Milne E. Increased intra-participant variability in children with autistic spectrum disorders: evidence from single-trial analysis of evoked EEG. Front Psychol. 2011. March 30;2:51. 10.3389/fpsyg.2011.00051 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref028] 28.Dinstein I, Heeger DJ, Lorenzi L, Minshew NJ, Malach R, Behrmann M. Unreliable evoked responses in autism. Neuron. 2012. September 20;75(6):981–91. 10.1016/j.neuron.2012.07.026 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref029] 29.Haigh SM, Heeger DJ, Dinstein I, Minshew N, Behrmann M. Cortical variability in the sensory-evoked response in autism. J Autism Dev Disord. 2015. May 1;45(5):1176–90. 10.1007/s10803-014-2276-6 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref030] 30.Zaidel A, Goin-Kochel RP, Angelaki DE. Self-motion perception in autism is compromised by visual noise but integrated optimally across multiple senses. Proc Natl Acad Sci. 2015. May 19;112(20):6461–6. 10.1073/pnas.1506582112 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref031] 31.Noel JP, Lakshminarasimhan KJ, Park H, Angelaki DE. Increased variability but intact integration during visual navigation in Autism Spectrum Disorder. Proc Natl Acad Sci. 2020. May 19;117(20):11158–66. 10.1073/pnas.2000216117 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref032] 32.Seriès P, Stocker AA, Simoncelli EP. Is the homunculus “aware” of sensory adaptation?. Neural Comput. 2009. December;21(12):3271–304. 10.1162/neco.2009.09-08-869 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref033] 33.Barlow HB. Possible principles underlying the transformation of sensory messages. Sensory Communication. 1961. September;1(01). [Google Scholar]

[pbio.3001215.ref034] 34.Młynarski WF, Hermundstad AM. Adaptive coding for dynamic sensory inference. Elife. 2018. July 10;7:e32055. 10.7554/eLife.32055 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref035] 35.Młynarski W, Hermundstad AM. Adaptability and efficiency in neural coding. bioRxiv. 2019. January 1:669200. [Google Scholar]

[pbio.3001215.ref036] 36.Simmons DR, Robertson AE, McKay LS, Toal E, McAleer P, Pollick FE. Vision in autism spectrum disorders. Vision Res 2009. November 10;49(22):2705–39. 10.1016/j.visres.2009.08.005 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref037] 37.Park WJ, Schauder KB, Zhang R, Bennetto L, Tadin D. High internal noise and poor external noise filtering characterize perception in autism spectrum disorder. Sci Rep. 2017. December 14;7(1):1–2. 10.1038/s41598-016-0028-x [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref038] 38.Lu ZL, Dosher BA. Characterizing observers using external noise and observer models: assessing internal representations with external noise. Psychol Rev. 2008. January;115(1):44. 10.1037/0033-295X.115.1.44 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref039] 39.Lu ZL, Dosher BA. External noise distinguishes attention mechanisms. Vision Res. 1998. May 1;38(9):1183–98. 10.1016/s0042-6989(97)00273-3 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref040] 40.Goldstone RL. Perceptual learning. Annu Rev Psychol. 1998. February;49(1):585–612. 10.1146/annurev.psych.49.1.585 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref041] 41.Maniglia M, Seitz AR. Towards a whole brain model of Perceptual Learning. Curr Opin Behav Sci. 2018. April 1;20:47–55. 10.1016/j.cobeha.2017.10.004 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref042] 42.Harris H, Israeli D, Minshew N, Bonneh Y, Heeger DJ, Behrmann M, et al. Perceptual learning in autism: over-specificity and possible remedies. Nat Neurosci. 2015;18 (11):1574–6. 10.1038/nn.4129 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref043] 43.Schoups A, Vogels R, Qian N, Orban G. Practising orientation identification improves orientation coding in V1 neurons. Nature. 2001. August;412(6846):549–53. 10.1038/35087601 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref044] 44.Ganguli D, Simoncelli EP. Efficient sensory encoding and Bayesian inference with heterogeneous neural populations. Neural Comput. 2014. October;26(10):2103–34. 10.1162/NECO_a_00638 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref045] 45.Clifford CW, Webster MA, Stanley GB, Stocker AA, Kohn A, Sharpee TO, et al. Visual adaptation: Neural, psychological and computational aspects. Vision Res. 2007. November 1;47(25):3125–31. 10.1016/j.visres.2007.08.023 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref046] 46.Cohen MR, Maunsell JH. Attention improves performance primarily by reducing interneuronal correlations. Nat Neurosci. 2009. December;12(12):1594. 10.1038/nn.2439 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref047] 47.Coen-Cagli R, Solomon SS. Relating divisive normalization to neuronal response variability. J Neurosci. 2019. September 11;39(37):7344–56. 10.1523/JNEUROSCI.0126-19.2019 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref048] 48.Carandini M, Heeger DJ. Normalization as a canonical neural computation. Nat Rev Neurosci. 2012. January;13(1):51–62. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref049] 49.Rosenberg A, Patterson JS, Angelaki DE. A computational perspective on autism. Proc Natl Acad Sci. 2015. July 28;112(30):9158–65. 10.1073/pnas.1510583112 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref050] 50.Li B, Peterson MR, Freeman RD. Oblique effect: a neural basis in the visual cortex. J Neurophysiol. 2003. July;90(1):204–17. 10.1152/jn.00954.2002 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref051] 51.Gowen E, Hamilton A. Motor abilities in autism: a review using a computational context. J Autism Dev Disord. 2013. February 1;43(2):323–44. 10.1007/s10803-012-1574-0 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref052] 52.Luu L, Stocker AA. Post-decision biases reveal a self-consistency principle in perceptual inference. Elife. 2018. May 22;7:e33334. 10.7554/eLife.33334 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref053] 53.Stocker AA, Simoncelli EP. Noise characteristics and prior expectations in human visual speed perception. Nat Neurosci. 2006. April;9(4):578–85. 10.1038/nn1669 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref054] 54.Karvelis P, Seitz AR, Lawrie SM, Seriès P. Autistic traits, but not schizotypy, predict increased weighting of sensory information in Bayesian visual integration. Elife. 2018. May 14;7:e34115. 10.7554/eLife.34115 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref055] 55.Coskun MA, Varghese L, Reddoch S, Castillo EM, Pearson DA, Loveland KA, Papanicolaou AC, Sheth BR. Increased response variability in autistic brains?. Neuroreport. 2009. November 25;20(17):1543–8. 10.1097/WNR.0b013e32833246b5 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref056] 56.Manning C, Tibber MS, Dakin SC. Visual integration of direction and orientation information in autistic children. Autism & Developmental Language Impairments. 2017. February;2:2396941517694626. [Google Scholar]

[pbio.3001215.ref057] 57.Manning C, Tibber MS, Charman T, Dakin SC, Pellicano E. Enhanced integration of motion information in children with autism. J Neurosci. 2015. May 6;35(18):6979–86. 10.1523/JNEUROSCI.4645-14.2015 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref058] 58.Butler JS, Molholm S, Andrade GN, Foxe JJ. An examination of the neural unreliability thesis of autism. Cereb Cortex. 2017. January 1;27(1):185–200. 10.1093/cercor/bhw375 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pbio.3001215.ref059] 59.Baron-Cohen S, Wheelwright S, Skinner R, Martin J, Clubley E. The autism-spectrum quotient (AQ): Evidence from asperger syndrome/high-functioning autism, malesand females, scientists and mathematicians. J Autism Dev Disord. 2001. February;31(1):5–17. 10.1023/a:1005653411471 [DOI] [PubMed] [Google Scholar]

[pbio.3001215.ref060] 60.Rutter M, Bailey A, Lord C. The social communication questionnaire: Manual. Western Psychological Services. 2003. [Google Scholar]

[pbio.3001215.ref061] 61.Efron B, Tibshirani RJ. An introduction to the bootstrap. CRC press; 1994. May 15. [Google Scholar]

PERMALINK

Individuals with autism spectrum disorder have altered visual encoding capacity

Jean-Paul Noel

Ling-Qi Zhang

Alan A Stocker

Dora E Angelaki

Roles

Abstract

Introduction

Fig 1. Theoretical and conceptual framework.

Results

Heightened variability and reduced learning in autism

Fig 2. Experimental protocol and individual participant performance.

Fig 3. Orientation perception in combined neurotypical and ASD participant.

Reduced capacity and aberrant allocation of encoding resources in autism

Fig 4. Quantification and parametrization of FI in neurotypical and ASD individuals.

Reduced initial encoding capacity is correlated with inflexibility in updating resource allocation

Fig 5. Correlation between sensory encoding capacity and flexibility in encoding resources allocation.

Discussion

Materials and methods

Participants

Materials and procedures

Data analysis

Supporting information

Acknowledgments

Abbreviations

Data Availability

Funding Statement

References

Decision Letter 0

Lucas Smith

Roles

Decision Letter 1

Lucas Smith

Roles

Author response to Decision Letter 1

Decision Letter 2

Lucas Smith

Roles

Author response to Decision Letter 2

Decision Letter 3

Lucas Smith

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases