Process reveals structure: How a network is traversed mediates expectations about its architecture.

Network science has emerged as a powerful tool through which we can study the higher-order architectural properties of the world around us. How human learners exploit this information remains an essential question. Here, we focus on the temporal constraints that govern such a process. Participants viewed a continuous sequence of images generated by three distinct walks on a modular network. Walks varied along two critical dimensions: their predictability and the density with which they sampled from communities of images. Learners exposed to walks that richly sampled from each community exhibited a sharp increase in processing time upon entry into a new community. This effect was eliminated in a highly regular walk that sampled exhaustively from images in short, successive cycles (i.e., that increasingly minimized uncertainty about the nature of upcoming stimuli). These results demonstrate that temporal organization plays an essential role in learners' sensitivity to the network architecture underlying sensory input.

Introduction

From classic computational models[1,2] to the exploration of human brain dynamics[3,4], network science and cognitive science continually inform and advance one another. In network science, one essential question concerns how real-word external systems, from social groups[5] to natural language[6], are structured. In cognitive science, network-based frameworks have been applied to uncover the internal architecture that gives rise to human behavior[7]. Here, we focus on a novel question that lies firmly at the intersection of network science and cognitive science, with far-reaching implications for the study of human learning: What constrains the extraction of higher-order properties from an external complex system? This question stands in contrast to related work from a computational modeling perspective, which offers a means of specifying the supportive mechanisms that give rise to human behaviors. Instead, we address a persistent gap in the convergence of network and cognitive science by examining whether adult participants’ sensitivity to externally derived graph structure is influenced by its temporal organization, or the specific order in which its edges are traversed. First, we show that an implicit, trial-by-trial measure of behavior reflects the network topology underlying a given stimulus environment, in this case a continuous series of visual events. Crucially, by systematically altering how the elements of a network unfold in time, we then demonstrate that human learners rely on the path through a network as they generate expectations about its structure.

Across domains, emerging evidence suggests that learners are sensitive to the network properties of their surroundings. Measures such as the clustering coefficient of a word (an index of how closely its phonological neighbors are related to one another) have been shown to predict how robustly that word is ultimately acquired[8]. Additional evidence indicates that the distance between objects in a network based on their physical features influences how early those objects are labeled by young children[9]. Outside of the language domain, adult learners have been shown to exploit higher-order temporal relations to segment continuous streams of images into visual events. In related prior work[10], participants were exposed to a sequence of images generated by a random walk on a network comprised of three groupings (communities) of densely interconnected nodes. Because each node had equivalent degree (number of edges incident to that node), a random walk on this network resulted in a continuous sequence of images with equivalent pairwise transition probabilities (TPs). In the statistical learning literature, peaks and dips in TPs constitute a powerful cue to event structure[11-13]; low TPs are considered an index of unpredictability, which in turn signals an event boundary. Nonetheless, even in the absence of this key source of information, learners were able to explicitly segment the stream into distinct communities after an extended training phase.

The afore-mentioned findings offer important insight into learners’ sensitivity to complex patterns over and above local associations between elements in their environment (for a detailed discussion of this topic see[14]). Though they certainly build on a long history of artificial grammar learning (AGL)[15], such recent developments in the learning literature ultimately answer a distinct set of questions. Canonical AGL studies typically investigate the extent to which learners are able to judge, following an extended training period, the grammaticality of symbol sequences generated by a random walk along an arbitrary finite-state grammar. While network-based approaches are increasingly used to examine which topological properties (e.g., community structure) influence learning, AGL paradigms are often used to examine the nature of learning separately from this sort of manipulation of (e.g., is it implicit/explicit[16]? Rule-based or chunk-based[17]?) Taken together, all network-based applications have great potential to answer fundamental questions about the scale at which learning operates.

However, if this line of inquiry is to continue to flourish, it is also essential to probe the process through which higher order patterns are revealed to the learner. While learning is clearly influenced by the topological properties underlying sensory input, the present study addresses whether learning is additionally influenced by the order (walk) in which the edges of a graph are traversed. Phrased another way, how does the distribution of information in time affect learners’ expectations about an underlying network architecture? This question is of particular importance as research on learners’ sensitivity to the topological properties of their surroundings continues to grow, mainly because the process through which information unfolds in naturalistic contexts (i.e., when processing language or visual events) is (often) clearly non-random.

A largely untapped method of examining the influence of walk structure is to continuously monitor learners’ processing speed as the elements of a graph-based sequence unfold (see also[18]). To the extent it varies by walk structure, an increase in reaction time (RT) associated with the unexpected (i.e., a surprisal effect) would serve as compelling evidence that learners are indeed sensitive to the process through which higher order patterns are revealed to them (this premise has been richly explored in the language domain[19]; for a discussion of surprisal in the context of syntactic comprehension see[20]). Though not necessarily framed in terms of expectation generation, related reaction time techniques have also been used to examine the acquisition of complex motor sequences[21], occasionally in the context of artificial grammar learning[22].

In the present study, we exposed adult learners to sequences of fractal images generated by three distinct walks on an identical underlying graph: a random walk, a walk consisting of successive Eulerian paths, and a walk consisting of successive Hamiltonian paths. In contrast to a random walk, Eulerian and Hamiltonian paths are highly structured, sampling exhaustively from all edges and nodes in a network, respectively (Fig. 1). The influence of temporal organization was examined using a graph with dense community structure[10]; nodes on that graph each represented a distinct visual image and edges connecting pairs of nodes represented their co-occurrence over a period of continuous exposure. Random, Hamiltonian, and Eulerian walk types were selected to probe two dimensions we hypothesized to be of great importance to the learner: uncertainty and redundancy. Because of the highly structured nature of the Eulerian and Hamiltonian paths, uncertainty was reduced relative to the Random condition (particularly rapidly for the Hamiltonian paths, which involved presentations of nodes in series of only 15). As each path progressed, the probability that a given node or edge would be encountered increased steadily (i.e., if it was not previously encountered in that series). Thus, as the set of possible traversals through the network was systematically narrowed, uncertainty about upcoming stimuli also decreased. On the other hand, we hypothesized that redundancy within a sequence might also constitute an important cue to event structure. In contrast to Hamiltonian paths, Eulerian paths and random walks tend to stay inside a given community, resulting in prolonged and repeated exposure to the common connections between nodes within a cluster. Exposure to these common connections, when presented in close temporal proximity, might then serve as a crucial learning cue.

Figure 1

Representation of the graph and walk structure underlying visual sequences. The graph consisted of three distinct communities of interconnected nodes (shown in yellow, teal, and purple). Each node in the graph corresponded to a unique fractal image, and edges between nodes corresponded to their possible co-occurrence in a sequence. Sequences were generated by “walking” along the edges of the graph randomly, or according to successive Eulerian and Hamiltonian paths. In the color-coded walk samples shown above, we illustrate that sequences generated by Random walks and Eulerian paths tended to stay within a given community (relative to Hamiltonian paths, which only sparsely sampled from each community).

During exposure to each walk type, participants completed a task requiring them to indicate via button press whether or not each image was rotated away from its canonical orientation. We then examined RTs to determine whether learners were developing community-based expectations about upcoming stimuli. Specifically, we contrasted transition nodes, which represented entry into a new community of images, with pre-transition nodes, the images that directly preceded them. If learners were successfully developing higher-order expectations, then we should observe an increase in processing time for the transition nodes. Below, we detail evidence that expectation- based processing indeed operates at the community level, but we implicate a crucial role for redundancy in extracting the topological properties of a densely clustered network. Compellingly, our findings extend beyond earlier work demonstrating that learners exploit graph structure to segment events, even in the absence of variations in transition probabilities[10]. Instead, we systematically vary walk structure in order to probe whether implicit expectations about underlying graph topology are influenced by the manner in which that topology unfolds from the earliest stages of learning onward.

Results

Rotation Detection

Participants generally excelled in distinguishing between the rotated and canonical fractal images during the exposure phase. Results indicate that cover task compliance was high across conditions (Fig. 2; mean A′ = 0.90, s.d. = 0.08; versus chance, t(59) = 90.88, p < 0.001).

Figure 2

Boxplots of reaction time increases across experimental conditions (N = 59). Cross-community surprisal effects were calculated by subtracting, for each participant, mean RTs for pre-transition nodes from mean RTs for transition nodes. A value greater than 0 indicates an increase in RT upon entry into a new community during the exposure phase. Note that strong evidence for surprisal is observed only for walk types involving repeated exposure to common connections within the same community (Eulerian and Random). No surprisal effect was observed for participants in the Hamiltonian condition.

Model 1: Random versus Hamiltonian Paths

In the first of a series of mixed effects models (1–3), RTs were regressed onto all main effects and interactions of Node Type (pre-transition versus transition), Condition (Random versus Hamiltonian or Random versus Eulerian) and Trial Number (continuous from 1–1395). To capture inter-individual variations, each model also included the fullest random effects structure that allowed the model to converge (for further detail, refer to Materials and Methods). When comparing RTs for the Random and Hamiltonian conditions (Model 1), we found significant main effects of Node Type (β = 10.85, t = 6.28, p < 0.001) and Trial (β = −31.27, t = −7.72, p < 0.001). Overall, RTs for pre-transition nodes were facilitated relative to transition nodes, and, as anticipated, participants generally sped up over the course of the experiment. Crucially, we observed a significant interaction between Node Type and Condition (β = −8.53, t = −4.94, p < 0.001), indicating that learners in the Random condition exhibited a significantly greater cross-community increase in processing time (referred to as a surprisal effect in Fig. 2) relative to learners in the Hamiltonian condition. This result is particularly impactful given that transition probabilities between pairs of nodes were equated in the Random condition. Finally, a significant interaction between Condition and Trial demonstrated that across both node types, learners were increasingly faster at processing elements in the Random condition (β = 8.28, t = 2.05, p = 0.048). Results for all main effects and their interactions are summarized in Table 1. Intriguingly, the Condition*Trial interaction was also maintained when examining all nodes in the network, as opposed to solely the boundary nodes (β = 11.35, t = 2.94, p = 0.005). Thus, despite the increase in processing times associated with transition nodes in the Random condition, overall RTs sped-up more quickly over time relative to the Hamiltonian condition (see also Supplementary Section 1).

Table 1

Coefficients (and corresponding t-values and p-values) for each predictor in a model examining the effect of Node Type (pre-transition versus transition), Condition (Model 1: Random versus Hamiltonian; Model 2: Random versus Eulerian; Model 3: Eulerian versus Hamiltonian), and Trial on RTs from the exposure phase. Significant values (determined using the Sattherwaite approximation and corresponding to p < 0.05) are bolded.

Predictor	Coefficient	T-value	P-value
MODEL I
Node Type (pre v. transition)	10.85	6.28	<0.001
Condition (Random v. Hamiltonian)	−5.96	−0.55	>0.250
Trial	−31.27	−7.72	<0.001
Node Type* Condition	−8.53	−4.94	<0.001
Node Type*Trial	−2.43	−1.54	0.123
Condition*Trial	8.28	2.05	0.048
Node Type*Condition*Trial	0.98	0.62	>0.250
MODEL II
Node Type (pre v. transition)	18.71	9.64	<0.001
Condition (Random v. Eulerian)	6.79	0.47	>0.250
Trial	−34.17	−8.89	<0.001
Node Type* Condition	−0.61	−0.32	>0.250
Node Type*Trial	−0.85	−0.48	>0.250
Condition*Trial	5.55	1.44	0.157
Node Type*Condition*Trial	2.49	1.40	0.163
MODEL III
Node Type (pre v. transition)	10.23	6.58	<0.001
Condition (Eulerian v. Hamiltonian)	−12.66	−1.02	>0.250
Trial	−25.37	−−5.60	<0.001
Node Type* Condition	−7.87	−5.06	<0.001
Node Type*Trial	0.19	0.12	>0.250
Condition*Trial	2.83	0.62	>0.250
Node Type*Condition*Trial	−1.59	−1.05	>0.250

Since Model 1 revealed a significant interaction between Node Type and Condition and Condition*Trial, a simple effects analysis was used to isolate the effects of Node Type and Trial separately for each condition. Results revealed that the Node Type*Condition interaction could be traced to a significant effect of Node Type exclusive to the Random condition (β = 19.38, t = 6.78, p < 0.001). No effect of Node type was observed for the Hamiltonian condition (β = 2.33, t = 1.20, p = 0.236). In contrast, the significant Condition*Trial interaction was not attributed to a null effect of Trial in the Hamiltonian condition (β = −22.99, t = −4.14, p = 0.002). Rather, the magnitude of the Trial effect was simply larger in the Random condition (β = −39.55, t = −6.71, p < 0.001).

Model 2: Random versus Eulerian Paths

Significant main effects of Node Type and Trial were maintained when examining the Random and Eulerian conditions. However, neither the Node Type* Condition interaction (β = −0.61, t = −0.32, p > 0.250) nor the Condition* Trial interaction (β = 5.55, t = 1.44, p = 0.157) were significant. Thus, neither the magnitude of the cross-community RT increase nor the overall speed-up of RTs in time differed between walk types that shared the property of redundancy. Results are summarized in Table 1. As we found no differences between the Eulerian and Random conditions, we did not perform a repetition priming analyses on Model 2 (see below).

Model 3: Eulerian versus Hamiltonian Paths

Significant main effects of Node Type and Trial were maintained when examining the Eulerian and Hamiltonian conditions (Table 1). Unsurprisingly given the results of Model 1, we also observed a significant interaction between Node Type and Condition (β = −7.87, t = −5.06, p < 0.001). However, we found no significant interaction between Node Type and Trial (β = −2.83, t = 0.62, p > 0.250).

Repetition Priming

Although we were interested in stimulus history effects as they relate to the network structures participants were learning, other types of stimulus history effects, such as item-specific repetition priming, could affect our observed measures. Processing times associated with a stimulus are known to decrease with recent exposure to that stimulus[23], which could explain a relative increase in RTs for transition relative to pre-transition nodes[10]. Under accounts that propose a distinction between expectation generation and priming effects[24], our results might then be said to reflect phenomena unrelated to learning. To be clear, repeated instances of the same node within a short time frame might in fact serve as an important part of the learning mechanisms driving expectations about community structure. In fact, while at a far more complex scale, evidence from natural language processing and production implicate a relationship between, e.g., syntactic priming and implicit learning[25-27]. Nonetheless, here we take a conservative approach in attempting to exclude low-level perceptual priming as the sole driver of pre-transition facilitation effects.

The decision to pursue this analysis was motivated in part by a related supplemental analysis from Schapiro et al. There, the authors compared orientation-judgment RTs for transition nodes relative to all other nodes for interleaved Random walks and Hamiltonian paths. We note that two features of this design, both addressed in the present work, make it challenging to interpret their corresponding RT results. First, the most direct comparison between conditions (Hamiltonian versus Random) occurred after learners had already been trained on 35-minutes of exposure to the Random condition. Second, interspersion of the two walk types could potentially influence the learning process, via carry-over effects, in unknown ways. Nonetheless, parallel to the present findings, RT increases for transition nodes were observed for random sequences but not Hamiltonian paths. However, the authors suggest that these RT effects can be attributed to priming, and, accordingly, that event segmentation can operate independently of how the nodes of a network unfold. Thus, despite clear similarities in overall results, their interpretation stands in contrast to the present proposal: that RT increases across community boundaries instead reflect learners’ expectations based on the temporal organization of topological patterns.

Admittedly, there are potentially multiple methods of assessing priming effects, none of which are likely to perfectly capture this phenomenon. Here, we focus on the inclusion of two confound predictors, which we have labeled Lag10 and Recency (for an alternative method of capturing priming, refer to Supplementary Section 2). These two predictors represented the number of times each image was seen in the previous 10 trials and the number of trials elapsed since each image was seen, respectively. After accounting for repetition priming effects in the Random versus Hamiltonian comparison, each of the significant main effects (Node Type: β = 6.26, t = 3.52, p < 0.001 and Trial: β = −31.39, t = −7.94, p < 0.001), as well as the interactions between Node Type and Condition (β = −4.40, t = −2.48, p = 0.014) and Condition and Trial (β = 8.33, t = 2.11, p = 0.041), were maintained. Thus, though participants clearly were sensitive to repetition priming effects (i.e., Lag10: β = −6.03, t = −3.00, p = 0.003; and Recency: β = 7.71, t = 3.23, p = 0.002), these confound predictors did not account fully for cross-community RT increases. After we confirmed that our significant interactions were maintained when accounting for priming, we verified via model comparison that this full model provided a superior fit to the data relative to a model that did not include main effects or interactions of Condition and Node Type (i.e., that only included both repetition priming predictors and Trial Number). Indeed, a comparison of the log likelihood ratio between the two models, after they were fit using a Maximum Likelihood procedure, was significant (χ2 = 19.66, p = 0.003). These results indicate that it is essential to account for walk type in evaluating cross-community RT increases.

Post-Exposure Measures

To make direct contact with related prior work[10], participants were also asked to complete both an on-line and off-line post-test following exposure to each walk type. Neither measure revealed evidence of learning, but this result is largely uninterpretable, as data quality was much lower for this portion of the experiment (for extensive discussion of this issue and detailed results from both tasks, see Supplementary Information Section 3 ).

Discussion

Implicit measures of processing speed during learning revealed that, without variations in pairwise probabilistic information, learners exhibited a sharp increase in RTs when transitioning out of a community of images. While we make no claim that the observed reaction time signatures indicate explicit internal representations of external network architecture, processing times clearly indicate sensitivity to the topological properties underlying the learning environment. Because this effect was maintained when accounting for image-specific repetition priming, it suggests that learners successfully generated expectations reflective of the meso-level (community) structure underlying a sequence of visual events. Interestingly, the absence of any observed interactions between Node Type and Trial suggests that these expectations emerged very early in exposure[28]. Furthermore, evidence for community-driven expectations was limited to the Random and Eulerian conditions, both of which involved rich and repeated exposure to common connections within the same community. This property, which we term redundancy, represents a higher-order property than image-specific repetition priming. When the traversal of intra-community edges was sparsely distributed in time, as in the Hamiltonian condition, learners did not exhibit the same increase in RT upon entry into a new community. Moreover, though the Hamiltonian paths were rigidly structured, we suggest that the nature of this structure actually minimized uncertainty to the detriment of the learner (for a related discussion of ordering effects in a different sort of learning context, see[29]). While Eulerian paths also increasingly minimized uncertainty, they did so more gradually, operating over series of 30 edge traversals as opposed to 15 nodes. Thus, the redundancy inherent to the Eulerian paths was likely sufficient to override these comparatively subtle decreases.

The Hamiltonian paths also diverged from the other two walk structures in a key way: because each 15-node path always traversed inter-community edges but reached only a subset of intra-community edges, they essentially constituted a weighted walk with unbalanced transition probabilities. An alternative interpretation of the observed pattern of results might then be that sharp differences in pairwise regularities (i.e., a weighted walk structure) can block learners’ sensitivity to certain higher-order topological patterns, at least as indexed by RT increases indicative of expectation generation. Relatedly, a second interpretation might be that variations in pairwise statistics are computed concurrently with meso-scale regularities such that learners’ local and higher-order expectations effectively cancel out, yielding uniform response times at community boundaries. Historically, event segmentation has been linked to the information conveyed by an unexpected or less probable stimulus[11,13], which in turn demands a processing cost. We thus consider it unlikely that learners here acquired knowledge of meso-scale architecture in the absence of observable cross-community RT increases. Nevertheless, the present data cannot rule out this possibility. While the sum of results presented here offer insight into relevant features of temporal organization (e.g., biased local statistics, time spent within a community, the increasing probability of encountering a given node), a crucial avenue of future research will be to probe precisely which features impede or facilitate learning, in addition to how those features might interact with more general cognitive processes (e.g., working memory capacity[30]).

To summarize, our results are powerful for a number of reasons. First, they indicate that, depending on the temporal organization of an externally-defined network, learners develop remarkably high-level (implicit) expectations about the nature of the upcoming signal. Compellingly, they do so even when pairwise transition probabilities between adjacent and non-adjacent elements are equated. The claim that expectation-based processes operate at a larger scale than local relationships accords with evidence from the sentence processing literature[19]. However, similar effects had not been investigated, particularly from a graph theoretical perspective, outside the language domain. Indeed, the properties we have presently examined through the manipulation of walk structure have ties to a much wider literature on language processing and production[20,31,32]. Extensive work on artificial grammar learning has probed the learning process using a graph-oriented framework (e.g., in the form of finite state grammars[15-17,22]), but has not explored the specific impact of the path used to generate grammatical sequences as related to on-line variations in processing time. Second, these findings shed light on a longstanding question surrounding the optimal learning paths in information networks[33]. A combination of computational modeling[34] and measures of semantic fluency[35] in humans indicate that retrieval processes operating on densely clustered semantic networks are optimized by algorithms that sample densely from those clusters before moving to another. Here, we have extended this work to the study of learning, offering evidence that redundancy in walk structure might also spur sensitivity to temporal communities in the context of continuous events. Thus, at least in a dynamic sensory environment, the fixed graph topology underlying incoming stimuli does not itself appear sufficient for learning. Rather, the order in which that topology is revealed to the learner can guide (or impede the formation of) expectations indicating the extraction of higher-order architectural properties. The extent to which computational models of the internal learning process (e.g., both prediction[22] and chunk-based frameworks[36]) can capture this externally-mediated phenomenon represents another open and exciting area of future research.

Materials and Methods

Participants

As verified by their assigned worker ID, 60 unique participants (20 per experimental condition) completed this study through Amazon Mechanical Turk, an on-line marketplace in which adult workers can perform behavioral experiments in exchange for financial compensation. Participants were paid at a rate of $0.08 per minute. To incentivize compliance with the task, they also received a completion bonus of $1.00 per phase of the experiment (exposure, segmentation, and odd-man out), and an additional $1.00 bonus if their cover task performance during the exposure phase exceeded 90% accuracy. Thus, total compensation for a participant who completed all phases of the experiment ranged from $7.00–$8.00. Participants communicated informed consent to engage in the study. Methods adhered to the guidelines and regulations of the Institutional Review Board (IRB) of the University of Pennsylvania. This committee (IRB at the University of Pennsylvania) also approved experimental protocols.

Materials

Stimuli consisted of 15 grayscale images generated via the Qbist filter in the GNU Image Manipulation program (v. 2.8.14; www.gimp.org). We elected to use unfamiliar, complex fractal images to minimize their nameability. Image-to-node assignment was randomized across participants.

Exposure sequence

In the Random condition, participants were exposed to a continuous sequence of 1400 images generated via a random walk on the graph shown in Fig. 1. This type of walk ensured that transition probabilities between pairs of nodes were equated, as each node in the graph was connected to precisely 4 other nodes. In the Hamiltonian condition, sequences were generated by concatenating a series of randomly selected forward and backward Hamiltonian paths, thereby ensuring that each of the 15 images was presented exactly once before a new path was initiated (1395 images total). The starting node of each path was a randomly selected node sharing an edge with the terminal node of the previous path. In the Eulerian condition, we concatenated a series of randomly selected forward and backward Eulerian paths (truncated at 1395 images total), each of which visited every edge of the graph exactly once before initiating a new path. Note that although the Hamiltonian and Eulerian conditions consisted of a concatenation of shorter paths, the resulting sequence was entirely continuous.

Procedure

Exposure phase

Prior to initiating the exposure phase, participants were instructed as follows: “you will see a stream of abstract images flashed on the screen one at a time. Your job is to keep your eyes on the screen as the stream progresses. Over time, parts of the stream may become familiar to you. This part will take around 35 minutes”. Participants were also informed that while they were viewing the images, they should indicate whether each image appeared in its canonical orientation (by pressing 1 on their keyboard) or whether it was rotated 90 degrees to the left (2 on their keyboard). Unknown to them, 15% of all images were rotated. To assist them on this task, participants first completed a study phase in which they viewed each image in its canonical orientation for 5 seconds. Next, they were asked to distinguish between a canonical and rotated image. Each trial was repeated until participants answered correctly. Furthermore, participants were instructed that they would hear a high-pitched tone if they responded incorrectly during the exposure phase and a low-pitch tone if they responded too slowly. Before beginning the exposure phase, participants had to pass a multiple-choice quiz that tested their knowledge of how that phase was structured. The full quiz was repeated until the participant achieved 100% accuracy on all questions. This step was intended to ensure high data quality, which is a potential concern for experiments conducted outside the laboratory[37]. To this end, participants were also informed that if they responded incorrectly (or failed to respond) to greater than 10 trials in a row, the experiment would terminate automatically. Participants were randomly assigned to one of two exposure lists per condition.

Post-exposure measures

After the exposure phase, participants completed two post-exposure tests: a segmentation task and an odd- man out judgment. These tests were intended to evaluate the explicit expression of graph structure knowledge. In the segmentation task, participants were asked only to press the space bar when they observed a natural breaking point in the stream of images (i.e., a new community). In the odd-man out judgment, participants were simultaneously presented with three images, two of which were drawn from the same community. They were asked to indicate which image “did not belong” with the others. As described in Supplementary Information Section 3, data quality was much lower (loss of > 20% of participants), making it difficult to interpret the observed absence of learning effects. Lowered data quality was presumably due to a combination of fatigue effects and a lack of performance-based financial incentives for either post-exposure measure. The sum of these results highlights the importance of including financial incentives, preferably on an orthogonal cover task, for all phases of on-line experimentation[37].

Analyses

Data exclusions

Prior to implementing the mixed effects models described below, we eliminated all incorrect trials from the exposure phase (8.9% total data loss) followed by trials in which the image was rotated away from its canonical orientation (a further 11.0% loss). Next, the following pre-determined data trimming steps were performed. First, we removed implausible RTs (i.e., less than 100 ms; 2.7% loss). In accord with common approaches in reading time studies, we then removed outlier data points from each subject constituting >3 standard deviations from their standardized mean processing times (1.5% data loss). We excluded all data from one subject from the Eulerian condition with an extremely anomalous RT pattern (i.e., the afore-mentioned data trimming techniques resulted in removal of 94.1% of that participant’s data). All significant findings reported here hold without these data trimming techniques.

Indexing cross-community expectations

We implemented a series of linear mixed effects models using the lmer() function (library lme4, v. 1.1–10) in R v. 3.2.2[38,39]. All data and corresponding analyses are available upon request. We focused on two types of community boundary nodes: transition nodes, which represented entry into a new community, and pre-transition nodes, which represented the node immediately prior to entry into a new community. If participants were indeed sensitive to the community structure of the network, then we should find a sharp increase in RT for the transition node relative to the pre-transition node. In rare cases involving a forward and backward traversal of the same edge (e.g., 5-6-5), we counted only the first of two transition nodes (in this case, 6).

Our primary analysis of interest was split into two models, each of which compared one of the structured walks (either Hamiltonian or Eulerian) to the Random condition. For completeness, we also present results from a third model comparing the Eulerian and Hamiltonian conditions. We adopted a multiple model approach because the inclusion of all three walk types in the same model resulted in excessive multicollinearity between effects of interest, even when predictors were centered (rs were > 0.6). RTs were regressed onto all main effects and interactions of Node Type (pre-transition versus transition), Condition (Random versus Hamiltonian or Random versus Eulerian) and Trial Number (1–1395). Predictors were centered to reduce multicollinearity. Each model also included the fullest random effects structure that allowed the model to converge. For Model 1 (Random versus Hamiltonian), Model 2 (Random versus Eulerian), and Model 3 (Eulerian versus Hamiltonian), this corresponded to a random intercept for participant and by-participant random slopes for Trial, Node Type, and the interaction between the two.

Accounting for repetition priming

To the extent that recency of stimulus repetitions varied by walk type, we sought to measure priming effects and to distinguish these effects from those attributable to the violation of expectations. We re-ran the model comparing the Random and Hamiltonian conditions, this time including two additional predictors intended to capture perceptual priming effects. The Lag10 predictor indexed the number of times participants had seen a given node in the previous 10 trials (range = 0–4; median = 0), while the Recency predictor indexed, for each node, the number of trials that elapsed since it was last seen (range 0–140; median 15). Thus, if participants viewed the sequence 1-2-5-4-3-5-3-1-3-2-5, the final node (5) would be assigned a Lag10 score of 2 and a Recency score of 5. To account for repetition priming effects, we regressed RTs from the Random and Hamiltonian conditions onto all main effects and interactions of Node Type (pre versus transition), Condition and Trial Number, plus the addition of these two priming predictors, Lag10 and Recency. The repetition priming model included a random intercept for participant and by-participant random slopes for Trial, Node Type, Recency, and the interaction between Trial and Node Type (the model failed to converge with the addition of the Lag10 predictor to the random effects term). In Supplementary Section 2, we describe an alternative method of accounting for priming effects and note that all original significant findings were maintained.