Chaos, Solitons & Fractals 44 (2011) 160–168

Contents lists available at ScienceDirect

Chaos, Solitons & Fractals Nonlinear Science, and Nonequilibrium and Complex Phenomena journal homepage:

Strong anticipation: Multifractal cascade dynamics modulate scaling in synchronization behaviors Damian G. Stephen a,⇑, James A. Dixon b,c a

Wyss Institute for Biologically Inspired Engineering, Harvard University, 3 Blackfan Circle, Floor 5, Boston, MA 02115, United States Department of Psychology, University of Connecticut, 406 Babbidge Rd., Unit 1020, Storrs, CT 06269-1020, United States c Haskins Laboratories, 300 George St., New Haven, CT 06511, United States b

a r t i c l e

i n f o

Article history: Received 22 April 2010 Accepted 21 January 2011 Available online 18 February 2011

a b s t r a c t Previous research on anticipatory behaviors has found that the fractal scaling of human behavior may attune to the fractal scaling of an unpredictable signal [Stephen DG, Stepp N, Dixon JA, Turvey MT. Strong anticipation: Sensitivity to long-range correlations in synchronization behavior. Physica A 2008;387:5271–8]. We propose to explain this attunement as a case of multifractal cascade dynamics [Schertzer D, Lovejoy S. Generalised scale invariance in turbulent phenomena. Physico-Chem Hydrodyn J 1985;6:623–5] in which perceptual-motor fluctuations are coordinated across multiple time scales. This account will serve to sharpen the contrast between strong and weak anticipation: whereas the former entails a sensitivity to the intermittent temporal structure of an unpredictable signal, the latter simply predicts sensitivity to an aggregate description of an unpredictable signal irrespective of actual sequence. We pursue this distinction through a reanalysis of Stephen et al.’s data by examining the relationship between the widths of singularity spectra for intertap interval time series and for each corresponding interonset interval time series. We find that the attunement of fractal scaling reported by Stephen et al. was not the trivial result of sensitivity to temporal structure in aggregate but reflected a subtle sensitivity to the coordination across multiple time scales of fluctuation in the unpredictable signal. Ó 2011 Elsevier Ltd. All rights reserved.

1. The problem of anticipation Anticipation is a major challenge for the ecological, biological, and cognitive sciences. In order to survive, an organism must act in anticipation of events, in accordance with the temporal structure of its environment. Anticipation is a central idea in the rapidly developing field of robotics [1,2], particularly with a view towards engineering robotic devices for medical treatments and prostheses [3,4]. Understanding the physics underpinning architectures of anticipatory systems is a crucial challenge to be resolved.

⇑ Corresponding author. E-mail address: [email protected] (D.G. Stephen). 0960-0779/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.chaos.2011.01.005

Traditionally, the strategy used in the cognitive sciences is to endow the organism with an internal model of the environment that permits extrapolation into the future [5,6]. In this formulation, the organism at time t will have constructed of a model based on prior experiences leading up to t. The organism interprets the past trajectory of the model to predict the future. This traditional description of anticipation is computationally heavy and relies upon the theory to bestow, not just an internal model, but also the constructive and interpretive powers needed to create and use internal models [5,7,8]. Dubois [9] has called this form of anticipation ‘‘weak’’ and has proposed a ‘‘strong’’ alternative that does not rely on internal models. Whereas weak anticipation places the burden of anticipation on an intelligent agent with constructive and interpretive powers, strong anticipation emphasizes the embedding of an

D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

organism within its environment. The embedding of the organism within its environment asserts lawful constraints upon both the actions of the organism and the environmental effects on those actions. Anticipation emerges as a lawful regularity of the organism–environment system [10]. 2. Interpretations of strong anticipation in empirical research 2.1. Negative mean asynchrony in anticipatory synchronization Empirical research in strong anticipation has differed according to initial interpretations of anticipation in the absence of an internal model. One view has aligned strong anticipation to anticipatory synchronization resulting from appropriate couplings between master and slave systems [10]. The time delays in the master system will lead the slave system to synchronize with future states of the master. Anticipatory synchronization has been found in empirical studies of physical [11–13] and biological systems [14,15]. Here the interpretation of anticipation has been that of a negative mean asynchrony between stimulus and response [16,17]. This interpretation of anticipation entails that an organisms experience of time unfolds at an extremely local scale. 2.2. Sensitivity to long-range temporal correlation An alternative approach interprets strong anticipation as a more global coordination between an organism and its environment [18]. Traditionally, weak anticipation in the present is best formalized as a recursion to past states (i.e., the use of the model for predictions of the future). Dubois [19] formalized strong anticipation in terms of hyperincursion. In a hyperincursion, the present state is determined by both the set of past states and the set of future states. At a first glance, hyperincursion might seem as though it flies in the face of a basic fundamental assumption of logical scientific thought, namely, that causes precede effects. However, hyperincursive logic need not entail that causes follow effects. Rather, it may be understood to express the multiple time scales on which anticipatory behavior unfolds. A variety of biological [20–23], meteorological [24,25], geological [26–28] and astrophysical [29,30] phenomena exhibit long-range correlations (e.g., 1/f fractal scaling) across a wide range of time scales [31]. We may understand hyperincursive logic as expressing the concurrent effects of short-range behavior and of long-range behavior on the same fluctuation at time t in instances of all these phenomena. Indeed, Dubois himself stressed the role of fractal scaling in hyperincursive logic [32]. Anticipation may seem distinctive or exceptional because it is most often cast as a function native to cognitive systems endowed with consciousness [6,33,34]. However, fluctuations in human cognitive behavior often exhibit 1/f fractal scaling [35–37], suggesting that anticipation may exemplify a more general tendency of natural systems to


fluctuate at many time scales at once [38,39]. For example, synchronization with an isochronous (i.e., periodic) metronome is long-range correlated, whether in terms of the asynchrony between the finger taps with metronome onsets [40,41] or the intervals between finger taps for continuation tapping once the metronome has been turned off [41–43]. That is, even though the isochronous metronome fluctuates at only one time scale, the timing behavior that emerges in response to it fluctuates at time scales far beyond the metronomes period. Indeed, the ability to respond proactively to a single metronome onset is only one small detail of anticipatory fluctuations underlying synchronization. Negative asynchrony to a single event may be the limiting case, often a feature of anticipatory behavior but not necessarily the general case. For instance, anticipatory though it may often seem, human behavior is plainly fraught with missed opportunities, failed predictions, and stubbed toes, and our impressive negative asynchronies may be deeply founded in our experience with our more humbling positive asynchronies. It may help theories of strong anticipation to explore the more global coordination between an organisms multiscale fluctuations and the multiscale fluctuations of its environment. Such a global coordination may be a rich substrate from which negative asynchrony might emerge. It is important to note where each of the three above approaches to anticipation can place its explanatory burden. Theories of weak anticipation place the explanatory burden on internal models [5]. The local approach to strong anticipation often places the explanatory burden on appropriate coupling between an organism and its environment [10]. In this manuscript, we will seek to place the explanatory burden of the global approach to strong anticipation on the multifractal cascades observable in perceptual-motor fluctuations. It is possible that multifractal cascades and appropriate coupling may entail one another, but for now, we turn a blind eye to negative asynchrony to better explore the potential of the global understanding of strong anticipation. 3. Strong anticipation in entrainment to chaotic metronomes Stephen et al. [18] tested this fractal interpretation of strong anticipation in a synchronization experiment. They sought to investigate synchronization behaviors under a circumstance in which prediction was highly unlikely. Participants were asked to attempt to synchronize finger taps with a chaotic metronome derived from the Ikeda [44,45] map that changes with each iteration according to:

x_ ¼ cx  b sin xs

for c ¼ 1; b ¼ 20;


where x is the current value and where xs is a previous value of x s iterations before the current value. The Ikeda map begins with a random seed and generates successive values xi based on the immediately previous value xi1 and another previous value xis at a delay of s. For constructing the chaotic metronomes, values of Eq. (1) were normalized to range from 1 to 1.5 and were then used as the interval in seconds between successive onsets of the metronome.


D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

Metronomes were designed to contain only 1000 onsets; the participants task lasted about 20 min. Metronomes were generated using various settings of s (i.e., s = 2, 5, and 10) to produce metronomes with varying ranges of memory. A new metronome was generated for each participant with each setting of s. Linear redundancy measures confirmed that predictability of these metronomes was close to nil. Using monofractal detrended fluctuation analysis (DFA; [46,47]), Stephen et al. computed the Hurst exponents for both the interonset interval time series for the metronomes and the intertap interval time series for the tapping signal produced by participants. The Hurst exponents of intertap intervals correlated with the Hurst exponents of interonset intervals, r = 0.96. Stephen et al. [18] ruled out the trivial interpretation of this finding, namely that finger taps matched onsets more or less accurately, with negligible, Gaussian-distributed asynchronies around the metronome onsets. Simulated tapping signals with Gaussian-distributed asynchronies failed to exhibit the same strong relationship between Hurst exponents. Stephen et al. concluded that human participants were not simply exhibiting a consistent or negligible asynchrony in their attempt to synchronize with the metronome and that, rather, even with their inconsistent asynchronies, human participants were all the more sensitive to the long-range temporal correlations of the metronome. This finding suggests that the long-range, multiscale fluctuations in human behavior may be strongly attuned to the long-range, multiscale fluctuations in an unpredictable environment (i.e., in this case, the metronome). Stephen et al. took this fractal attunement between organism and environment to be an example of strong anticipation, the coordination with temporal structure in the absence of local prediction.

4. Multifractal account of strong anticipation What remains to be explained is the means by which such strong attunement between behavior and environment might develop. The crucial question is how an organism could tailor its fluctuations to those presented in the experimental environment. We seek to explain this tailoring of fluctuations as the product of multifractality. Whereas purely monofractal dynamics consist of homogeneously self-similar cascades of fluctuations for all time scales and all sizes of fluctuations, multifractal dynamics hold when the self-similar cascades exhibit different scaling relationships for different scales and for different sizes of fluctuations [48–52]. Heterogeneity across time makes multifractality an ideal formalism for discussing the development of a new fractal scaling. Schertzer and Lovejoy [53] and colleagues [50,51] have developed Kolmogorov’s [54] notion of cascade dynamics into a theory of the development of multifractal cascades. In their model, fluctuations of relatively larger size cascade down upon fluctuations of relatively smaller sizes. Essentially, for any raw observable u(t) (e.g., a time series of individual intervals, whether between taps or between onsets), as we sample the observable over progressively larger scales L, the summed values of u(t) in each sample will, on average, represent a progres-

sively greater proportion P(L) of the summed values over the entire time series such that:

PðLÞ  La :


The foregoing singular relationship may exhaustively describe the distribution of u(t) through samples of scale L for a homogeneous system, but in a heterogeneous system, different regions i may exhibit diverse singular relationships,

Pi ðLÞ  Lai :


Adequate specification of such a heterogeneous system requires a spectrum of singularities (i.e., the multifractal spectrum). The presence of multiple singularities is evidence of multiplicative cascades in the flow of energy through the observable u(t), a major source of intermittent temporal structure. The width of the singularity spectrum (i.e., amax  amin) reflects the strength of these multiplicative cascades [55,56]. The significance of multiplicative cascades for our present purpose is that they entail a coordination of fluctuations (i.e., information and energy flow) among multiple time scales. The hyperincursive logic underpinning strong anticipation suggests that just such a coordination among multiple time scales may support the apparently predictive aspects of behavior without requiring an internal model. Although Stephen et al. [18] recognized the possibility that human behavior can attune to long-range structure, they failed to adequately address the possibility that synchronization behaviors might reflect a coordination among multiple scales. Importantly, the Hurst exponent estimated by monofractal analysis is insensitive to heterogeneity and hence to deeper multiplicative cascade structure [57,55]. The Hurst exponent speaks only to the structure of autocorrelation of a time series u(t) in aggregate, at a level of description statistically indistinguishable from linear forecasting methods (e.g., autoregressive fractionally integrated moving average [ARFIMA]; [58]) [59]. Hence, fractal scaling in entrainment with periodic metronomes has already been proposed to fold naturally into a traditional weak-anticipatory framework (e.g., [60,61]) in which linear phase correction adjusts the expression of a timekeeping model according to lag-1 asynchronies [62]. In this light, there need be no reason that the results of Stephen et al. do not reflect the trivial mapping of the autocorrelation, and the close relationship between Hurst exponents of interonset intervals and Hurst exponents of intertap intervals need not reflect anything about the coordination of synchronization behavior at multiple time scales. For instance, the same intertap intervals in entirely reversed order would have given rise to the same Hurst exponents but might not have reflected any sensitivity to the temporal structure of the interonset intervals of the chaotic metronome. We return to the data from [18] with the purpose of determining whether the results reported simply reflected weak-anticipatory sensitivity to overall autocorrelational structure or whether the results reported reflect the multiplicative cascades coordinating energy flow across multiple time scales in perceptual-motor fluctuations. In the


D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

latter case, multifractal cascade dynamics [53,50,51,63] would serve as a crucial foundation for the organization of synchronization behaviors. Multiplicative cascades coordinating multiple time scales have been documented in perceptual-motor fluctuations during a variety of simple cognitive and motor tasks [57], but it has not yet been made clear what practical role such cascade dynamics might have for behavior. The present work proposes to determine whether multifractal cascade dynamics might support strong anticipation. In what follows, we reanalyzed the intertap and interonset time series from [18] with Chhabra and Jensen’s [64,65] method for direct determination of the f(a) singularity spectrum. Stephen et al. [18] had drawn on detrended fluctuation analysis (DFA; [46,47]) in order to estimate Hurst exponents, but the multifractal generalization of DFA [52] depends on Legendre transformation for determining the singularity spectrum and can be unstable for short time series [57]. We estimated the width of the singularity spectrum (i.e., amax  amin) for each original time series as well as for a distribution of 50 phase-randomized surrogates for each original time series. These phase-randomized surrogates were constructed according to the iterative amplitude-adjusted Fourier-transform (IAAFT) algorithm [66]. For a given time series u(t), the IAAFT algorithm generates novel sequences of numbers reflecting the same probability density function (pdf) and the same power spectrum as the original time series. Hence, amax  amin for an IAAFT surrogate indicated the degree of multifractality to be expected simply as a by-product of the pdf and the power spectrum; the randomization of phase and consequent destruction of sequence entailed that amax  amin for an IAAFT surrogate did not bear information about the strength of a multiplicative cascade across multiple time scales structuring the actual sequence. The IAAFT surrogate bears the same overall autocorrelational structure of the original time series without sharing the actual sequence. A departure of amax  amin for the original time series from the distribution of amax  amin for the IAAFT surrogates would have indicated a role for multiplicative cascade across multiple time scales. The contrast between amax  amin for IAAFT surrogates and amax  amin for the original time series served for a contrast between expectations of the temporal structure from the weak- and strong-anticipatory views. If intertap interval time series exhibited amax  amin not significantly different from their corresponding IAAFT surrogates, then this evidence would be consistent with the weak-anticipatory notion that synchronization behavior simply tracks overall autocorrelational structure and that the sensitivity to long-range structure noted in [18] did not reflect any specific attunement to actual temporal sequence. On the other hand, the predictions from the strong-anticipatory view were, first, that the intertap interval time series would exhibit a amax  amin significantly different from that of their IAAFT surrogates and, second, that amax  amin of the intertap intervals should reflect the amax  amin of the corresponding interonset intervals. That is, weak anticipation in synchronization under unpredictable conditions should be indistinguishable from simply mimicking the

aggregate temporal structure. Strong anticipation in synchronization under unpredictable conditions should entail that, far from being a rough approximation of aggregate temporal, the coordination of multiple time scales in behavior should attune itself specifically to the coordination of multiple time scales in the environment. In this sense, whereas weak anticipation might simply be the tailoring of behavior to a strictly linear description of the environment, strong anticipation should be the subtle sensitivity of behavior to the multiplicative cascades structuring events in the environment. 5. Method Seven participants were instructed to attempt synchronization with their metronomes. Metronomes were generated by normalizing values of x from Eq. (1) (with parameters c, b, and s noted above) to produce onset intervals ranging from 1 to 1.5 s. At the beginning of each session, each participant received a new metronome, begun from a different random seed. Each metronome produced 1000 onsets. Metronomes were presented on a computer and, at each onset, produced a visual presentation on the computer screen (i.e., a white square appeared at the center of a blank screen). Participants were instructed to tap any key on the computer keyboard in synchrony with each new onset, to the best of his or her ability. Participants completed 1 to 2 sessions of this task for each setting of s. 6. Data analysis 6.1. Direct determination of the singularity spectrum Direct determination of the singularity spectrum [64,65] for a time series u(t) begins with calculating the probabilities Pi(L) for each ith bin of scale L.

PiL Pi ðLÞ ¼

k¼ði1ÞLþ1 uðkÞ



 Lai :


Next, the probabilities Pi(L) allow derivation of the partition function li(q, L) for each of N bins of scale L.

½Pi ðLÞq

li ðq; LÞ ¼ PN

i¼1 ½P i ðLÞ




where q is a statistical moment intended to leave probabilities unchanged for q = 1 and to emphasize higher or lower probabilities for q > 1 or q < 1, respectively. Hence, the partition function l(q, L) serves to statistically highlight the different densities of events across each of the N bins for a given scale L. The Hausdorff dimension f(q) of partition function l(q) is derived as the scaling relationship of the Shannon entropy of li(q, L) with scale L.

f ðqÞ ¼  lim


N 1 X l ðq; LÞ ln½li ðq; LÞ ln N i¼1 i

PN ¼ lim L!0


li ðq; LÞ ln½li ðq; LÞ ln L



The corresponding singularity strength a for the corresponding q is defined as


D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

aðqÞ ¼  lim


N 1 X l ðq; LÞ ln½Pi ðq; LÞ ln N i¼1 i

PN ¼ lim L!0


li ðq; LÞ ln½Pi ðq; LÞ ln L



The value a(q) is essentially the average of the singularity strengths ai for each ith bin for a given value of q. For values of q giving rise to strongly linear scaling relationships in Eqs. (6) and (7), when f(q) and a(q) form a singlehumped curve f(a), the time series u(t) is multifractal [67]. The f(a) curve is the singularity spectrum, and its width signifies the degree of multifractality [57,48]. 7. Results Example time series for interonset intervals in the metronome for an example participant and for intertap intervals in the corresponding example participants tapping are depicted in Figs. 1 and 2, respectively. We submitted each time series to the multifractal analysis described in the Methods section, including only those values of f(q) and a(q) for which the relationships outlined

in Eqs. (6) and (7) both exhibited a correlation coefficient r P 0.95. The q parameter was thus incremented and decremented from q = 1 by 0.5 until these scaling relationships weakened to below the r P 0.95 criterion. Fig. 3 shows the resulting multifractal (i.e., singularity) spectra for intertap intervals from an example participant and for the interonset intervals from the corresponding metronome. We were interested in two main features of the multifractality in the intertap intervals. First, we were interested in investigating whether, for the intertap interval time series, the width of the singularity spectrum amax  amin was not a trivial consequence of the pdf and power spectrum of the time series but rather an index of the coordination across multiple time scales in the original sequence. Second, we were curious whether the amax  amin for the intertap interval time series was sensitive to the amax  amin for the interonset interval time series. We now consider both points in the subsections that follow. 7.1. Comparison to surrogates To address the first point, we compared the singularity spectra for each intertap interval time series with a distri-

Fig. 1. Example time series of interonset-interval durations for a single chaotic metronome.

Fig. 2. Example time series of intertap-interval durations for a single participants entrainment to the metronome whose interonset intervals are illustrated in Fig. 1.

D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

bution of singularity spectra for 50 IAAFT surrogates (i.e., a new distribution of 50 spectra for each time intertap interval time series). To safeguard against the possibility that the pdf of the time series might spuriously widen the singularity spectrum irrespective of power spectrum, we also compared the singularity spectrum for each intertap interval time series with a distribution of singularity spectra for 50 shuffled copies (i.e., randomized reorderings) of the original time series (i.e., again, a new distribution of 50 spectra for each time intertap interval time series). We evaluated the differences between amax  amin for each original time series and amax  amin for each corresponding comparison distribution as one-sample t statistics. Throughout, the width of the singularity spectrum amax  amin for the original time series were significantly


different from that of both the shuffled copies and the IAAFT surrogates, jtsj ranging from 2.53 to 213.55 and hence above the criterion of 1.96 for significance at p < .05 (see Fig. 4). The only exception was one intertap time series whose amax  amin was not significantly different from that of its IAAFT surrogate, t = 0.15, p = 0.44. However, though almost always significantly different, singularity-spectra for the original time series were neither generically wider nor generically narrower than those for surrogates. Previous research has documented the presence of multiplicative cascades and, hence, the coordination of fluctuations across multiple time scales by demonstrating that, in response tasks, perceptual-motor fluctuations exhibited wider singularity spectra (i.e., greater amax  amin) than their corresponding shuffled copies

Fig. 3. Multifractal (i.e., singularity) spectra for intertap intervals from an example participant (circles) and for the interonset intervals from the corresponding metronome (triangles).

Fig. 4. t statistics for singularity-spectrum widths for intertap interval time series when compared to the distribution of singularity-spectrum widths for shuffled copies (circles) and for IAAFT surrogates (triangles).


D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

Table 1 Coefficients from the multiple linear regression model of amax  amin for intertap intervals. Predictor





Intercept MetOrig MetIAAFT TapIAAFT MetOrig  MetIAAFT

0.04 1.72 2.23 0.09 19.58

0.02 0.54 1.37 0.09 37.86

1.71 3.17 1.63 1.02 0.52

0.11 <.01 0.13 0.33 0.61

and IAAFT surrogates [57]. It remained to be determined whether multifractality had bearing upon strong anticipation and, specifically, whether the changes in amax  amin for intertap intervals reflected sensitivity to changes in amax  amin for interonset intervals.

7.2. Relationships between multifractality of intertap intervals and multifractality of interonset intervals In order to test whether amax  amin for intertap intervals reflected sensitivity to amax  amin for interonset intervals, we ran a multiple linear regression model of amax  amin for intertap intervals. In order to make a compelling case for strong anticipation, we aimed to test the effect of amax  amin for interonset intervals on amax  amin for intertap intervals while also testing the effect of temporal structure in aggregate. Recall that a weak-anticipatory view would expect that intertap intervals should both be sensitive to temporal structure in aggregate and be organized according to temporal structure in aggregate. That is, it will be important not only to test an effect of amax  amin for interonset intervals on amax  amin on intertap intervals but also the effects of amax  amin for the IAAFT surrogates for the interonset intervals as well as for the intertap intervals. Whereas amax  amin for interon-

set intervals should be significant if the degree of coordination across multiple time scales in behavior responds to the degree of coordination across multiple time scales in the stimulus, the latter two effects should be significant if synchronization behaviors are temporally structured only in aggregate. The predictors chosen for this model were as follows: amax  amin for interonset intervals (MetOrig), mean amax  amin for the IAAFT surrogates of the interonset intervals (MetIAAFT), and mean amax  amin for the IAAFT surrogates of the intertap intervals (TapIAAFT). Because amax  amin for interonset intervals and mean amax  amin for the IAAFT surrogates of the interonset intervals correlated at r(20) = 0.89, p < .001, it was also important to include the interaction of these two singularityspectrum widths in order to judiciously estimate the effect of each width alone. Table 1 shows the coefficients returned by the multiple linear regression model of amax  amin for intertap intervals as planned above. There is only a main effect of amax  amin for the interonset intervals (B = 1.72, SE = 0.54, p < 0.01). No other effects significantly predict changes in amax  amin for intertap intervals. Model predictions of singularityspectrum width correlated with actual singularity-spectrum width, r = 0.98 (see Fig. 5). 8. Anticipation and its relationship to multifractal scaling The results of the present research point to the beginnings of an alternative explanation for anticipation. Instead of prediction from an internal model, anticipation may be the product of multifractal cascade dynamics [53,50, 51,63]. Changes in the long-range, multiscale fluctuations of human behavior are the result of fluctuations of many different sizes. In the present reanalysis, the Hurst exponents reported in [18] do not simply appear to have been

Fig. 5. Model predictions for the multiple linear regression of amax  amin for intertap intervals (vertical axis) plotted against actual amax  amin for intertap intervals (horizontal axis).

D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

a trivial matching the aggregate temporal structure. Rather, through an inquiry into the multifractal structure of both the metronomes and the tapping signals, the present reanalysis indicates that the tapping reflected a coordination across multiple time scales that closely matched that found in the metronomes. The multifractality of a given stimulus appears to modulate the multifractality of perceptual-motor fluctuations, suggesting a more subtle sensitivity to the intermittent temporal structure than the cognitive system would be able to glean if it simply consulted the power spectrum. 8.1. Implications for anticipation: weak and strong, local and global We had begun by considering three different approaches to anticipation: weak anticipation, strong anticipation on a temporally local scale, and strong anticipation on a temporally global scale. First, this research serves as a challenge for theories of weak anticipation, i.e., the notion that anticipation is prediction from an internal model [5,6]. Stephen et al. [18] had demonstrated that organisms may coordinate even to an unpredictable temporal structure. These results do not disprove the presence or function of internal models supporting an organism’s behavior. Rather, they suggest that understanding anticipation may not require so much recourse to an internal model, hidden deep within the architecture of anticipatory systems and invested with mysterious powers. These results suggest that the dynamics of anticipation may emerge from the sensitivity of the multifractal fluctuations of perceptual-motor coordination, at the outermost boundaries of the organism, to the multifractal structure in the environment. This finding may be aligned with closely with existing research on the role of embodiment in the development of anticipation among other cognitive capacities [68]. The present results provide new support for theories of strong anticipation. These results may serve to complement existing research on strong anticipation at a temporally local scale (e.g., [10]). Anticipation may not be so much predictive extrapolation of the present into the future as a reactive, feedforward coordination of preexisting fluctuations of very many sizes across multiple time scales. Indeed, one general finding of research in anticipatory synchronization has been that apparent prediction in the long term may ultimately be the product of reactivity in the short term [11–13]. In the longer term, we have found that strong anticipation may depend on the coordination of perceptual-motor fluctuations across multiple time scales and the multiplicative cascades that such coordination entails and on the capacity of this coordination to adapt to the environmental constraints. Because we have shown that anticipatory behavior is not specifically sensitive to aggregate temporal structure, the notion of strong anticipation at a ‘‘temporally global scale’’ is misleading insofar as the intertap intervals exhibited sensitivity to the coordination across multiple time scales. We conclude that strong anticipation unfolds at very many time scales according to multiplicative cascades and hence that strong anticipation is a phenomenon of multifractal dynamics.


This research represents the latest in an attempt to understand cognitive performance as emerging from the fractal and multifractal dynamics of complex physical systems [69–72,57]. Further research may address the generality of these findings, both to multifractal systems not typically designated as cognitive and to cognitive systems in a wider variety of task domains.

Acknowledgments Partial support was provided by NSF Grant No. BCS0643271 to J.D.

References [1] Hoffmann H. Perception through visuomotor anticipation in a mobile robot. Neural Networks 2007;20:22–3. [2] Sporns O, Alexander WH. Neuromodulation and plasticity in an autonomous robot. Neural Networks 2002;15:761–74. [3] Doulgeri Z, Fasoulas J, Arimoto S. Feedback control for object manipulation by a pair of soft tip fingers. Robotica 2002;20:1–11. [4] Nicolelis MAL. Actions from thoughts. Nature 2001;409:403–7. [5] Rosen R. Anticipatory systems. New York: Pergamon Press; 1985. [6] Pezzulo G. Coordinating with the future: the anticipatory nature of representation. Mind Mach 2008;18:179–225. [7] Harnard S. The symbol grounding problem. Physica D 1990;42:335–46. [8] Searle JR. Minds, brains, and programs. Behav Brain Sci 1980;3:417–57. [9] Dubois DM. Mathematical foundations of discrete and functional systems with strong and weak anticipations. In: Butz MV, Olivier S, Gérard P, editors. Anticipatory behavior in adapative learning systems. LNAI, vol. 2684. Berlin: Springer-Verlag; 2003. p. 110–32. [10] Stepp N, Turvey MT. On strong anticipation. Cogn System Res 2010;11:148–64. [11] Voss HU. Dynamical long-term anticipation of chaotic states. Phys Rev Lett 2001;87:014102. [12] Voss HU. Real-time anticipating of chaotic states of an electronic circuit. Int J Bifurc Chaos 2002;12:1619–25. [13] Sivaprakasam S, Shahverdiev EM, Spencer PS, Shore KA. Experimental demonstration of anticipating synchronization in chaotic semiconductor lasers with optical feedback. Phys Rev Lett 2001;87:154101. [14] Toral R, Masoller C, Mirasso CR, Ciszak M, Calvo O. Characterization of the anticipated synchronization regime in the coupled Fitzhugh– Nagumo model for neurons. Physica A 2003;325:192–8. [15] Stepp N, Turvey MT. Anticipating synchronization as an alternative to the internal model. Behav Brain Sci 2008;31:216–7. [16] Radill T, Mates J, Ilmberger J, Pöppel E. Stimulus anticipation in following rhythmic acoustical patterns by tapping. Experientia 1990;46:762. [17] Repp B. Sensorimotor synchronization: a review of the tapping literature. Psychon Bull Rev 2005;12:969–92. [18] Stephen DG, Stepp N, Dixon JA, Turvey MT. Strong anticipation: sensitivity to long-range correlations in synchronization behavior. Physica A 2008;387:5271–8. [19] D.M. Dubois, Incursive and hyperincursive systems, fractal machine and anticipatory logic. In: Dubois DM, editors, Computing anticipatory systems: CASYS, Fourth int conf, AIP conf proc, vol. 573, Woodbury; 2001. p. 437–51. [20] Bassingthwaighte JB, Liebovitch LS, West BJ. Fractal physiology. New York: Oxford University Press; 1994. [21] Werner G. Fractals in the nervous system: conceptual implications for theoretical neuroscience. Front Physiol 2010;1:1–20. [22] West BJ. Where medicine went wrong: rediscovering the path to complexity. Hackensack, NJ: World Scientific; 2006. [23] West GB, Brown JH, Enquist BJ. A general model for the origin of allometric scaling laws in biology. Science 1997;276:122–6. [24] Eichner JF, Koscielny-Bunde E, Bunde A, Havlin S, Schellnhuber H-J. Power-law persistence and trends in the atmosphere: a detailed study of long temperature records. Phys Rev E 2003;68:046133. [25] Lovejoy S, Schertzer D, Allaire VC. The remarkable wide range spatial scaling of TRMM precipitation. Atmos Res 2008;90:10–32.


D.G. Stephen, J.A. Dixon / Chaos, Solitons & Fractals 44 (2011) 160–168

[26] Baker L, Giancola AJ, Allahdadi F. Fracture and ejecta spall mass distribution: lognormal and multifractal distributions. J Appl Phys 1992;72:2724–31. [27] Kirchner JW, Feng X, Neal C. Fractal stream chemistry and its implications for contaminant transport in catchments. Nature 2000;403:524–7. [28] Turcotte DL. Fractals and chaos in geology and geophysics. Cambridge University Press; 1997. [29] Hnat B, Chapman SC, Rowlands G. Intermittency, scaling, and the Fokker–Planck approach to fluctuations of the solar wind bulk plasma parameters as seen by the WIND spacecraft. Phys Rev E 2003;67:056404. [30] Iovane G, Laserra E, Tortoriello FS. Stochastic self-similar and fractal universe. Chaos Soliton Fract 2004;20:415–26. [31] Mandelbrot BB. The fractal geometry of nature. San Francisco, CA: Freeman; 1982. [32] Dubois DM. The fractal machine. Liège: Liège University Press; 1992. [33] Riegler A. Decision-making and anticipations: why Buridani’s ass is an unlikely creature. In: Smit I, Wallach W, Lasker GE, editors. Cognitive, emotive, and ethical aspects of decision making in humans and in artificial intelligence, vol. 4. Windsor: International Institute for Advanced Studies in Systems Research and Cybernetics; 2005. p. 35–40. [34] Rowe RM, McKenna FP. Skilled anticipation in real-world tasks: measurement of attentional demands in the domain of tennis. J Exp Psychol Appl 2001;7:60–70. [35] Gilden DL. Cognitive emissions of 1/f noise. Psychol Rev 2001;108:33–56. [36] Thornton TL, Gilden DL. Provenance of correlations in psychological data. Psychon Bull Rev 2005;12:409–41. [37] Kello CT, Anderson GG, Holden JG, Van Orden GC. The pervasiveness of 1/f scaling in speech reflects the metastable basis of cognition. Cogn Sci 2008;32:1217–31. [38] Treffner PJ, Kelso JAS. Dynamic encounters: long memory during functional stabilization. Ecol Psychol 1999;11:103–37. [39] Van Orden GC, Holden JG, Turvey MT. Self-organization of cognitive performance. J Exp Psychol Gen 2003;132:331–50. [40] Chen Y, Ding M, Kelso JAS. Long memory processes (1/fa type) in human coordination. Phys Rev Lett 1997;79:4501–4. [41] Roberts S, Eykholt R, Thaut MH. Analysis of correlations and search for evidence of deterministic chaos in rhythmic motor control by the human brain. Phys Rev E 2000;62:2597–607. [42] Delignières D, Lemoine L, Torre K. Time intervals production in tapping and oscillatory motion. Hum Movement Sci 2004;23:87–103. [43] Lemoine L, Torre K, Delignières D. Testing for the presence of 1/f noise in continuation tapping data. Can J Exp Psychol 2006;60:247–57. [44] Ikeda K, Matsumoto K. High-dimensional chaotic behavior in systems with time-delayed feedback. Physica D 1987;29:223–35. [45] Ikeda K, Daido H, Akimoto O. Optical turbulence: chaotic behavior of transmitted light from a ring cavity. Phys Rev Lett 1980;45:709–12. [46] Peng C-K, Buldyrev SV, Havlin S, Simons M, Stanley HE, Goldberger AL. Mosaic organization of DNA nucleotides. Phys Rev E 1994;49:1685–9. [47] Peng C-K, Havlin S, Stanley HE, Goldberger AL. Quantification of scaling exponents and crossover phenomena in nonstationary heartbeat time series. Chaos 1995;5:82–7. [48] Ivanov PC, Amaral LAN, Goldberger AL, Havlin S, Rosenblum MG, Stanley HE, et al. From 1/f noise to multifractal cascades in heartbeat dynamics. Chaos 2001;11:641–52. [49] Mandelbrot BB. Multiplications aléatoires itérées et distributions invariantes par moyenne pondérée aléatoire. CR Acad Sci Paris Ser A 1974;278:289–92.

[50] Schertzer D, Lovejoy S, Schmitt F, Chigirinskaya Y, Marsan D. Multifractal cascade dynamics and turbulent intermittency. Fractals 1998;5:427–71. [51] Seuront L, Schmitt F, Lagadeuc Y, Schertzer D, Lovejoy S. Universal multifractal analysis as a tool to characterize multiscale intermittent patterns: example of phytoplankton distribution in turbulent coastal waters. J Plankton Res 1999;21:877–922. [52] Kantelhardt JW, Zschiegner SA, Koscielny-Bunde E, Havlin S, Bunde A, Stanley HE. Multifractal detrended fluctuation analysis of nonstationary time series. Physica A 2002;316:87–114. [53] Schertzer D, Lovejoy S. Generalised scale invariance in turbulent phenomena. Physico-Chem Hydrodyn J 1985;6:623–35. [54] Kolmogorov AN. A local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers. Dokl Acad Nauk SSSR 1941;30:299–303. [55] Mandelbrot BB. Fractals and scaling in finance. New York: SpringerVerlag; 1997. [56] Halsey TC, Jensen MH, Kadanoff LP, Procaccia I, Shraiman BI. Fractal measures and their singularities: the characterization of strange sets. Phys Rev A 1986;33:1141–51. [57] Ihlen EAF, Vereijken B. Interaction-dominant dynamics in human cognition: beyond 1/fa fluctuation. J Exp Psychol Gen 2010;139:436–63. [58] Granger CWJ, Joyeux R. An introduction to long-memory models and fractional differencing. J Time Series Anal 1980;1:15–29. [59] Torre K, Delignières D, Lemoine L. Detection of long-range dependence and estimation of fractal exponents through ARFIMA modeling. Brit J Math Stat Psychol 2007;60:85–106. [60] Wing AM, Kristofferson AB. Response delays and the timing of discrete motor responses. Percept Psychophys 1973;14:5–12. [61] Vorberg D, Wing AM. Modeling variability and dependence in timing. In: Keele S, Heuer H, editors. Handbook of perception and action, vol. 2. New York: Academic Press; 1996. p. 181–262. [62] Delignières D, Torre K, Lemoine L. Long-range correlation in synchronization and syncopation tapping: a linear phase correction model. PLoS ONE 2009;4:e7822. [63] Schertzer D, Lovejoy S. Uncertainty and predictability in geophysics: chaos and multifractal insights. In: Sparks RSJ, Hawkesworth CJ, editors. State of the planet, frontiers and challenges in geophysics. American Geophysical Union; 2004. p. 317–34. [64] Chhabra AB, Jensen RV. Direct determination of the f(a) singularity spectrum. Phys Rev Lett 1989;62:1327–30. [65] Chhabra AB, Meneveau C, Jensen RV, Sreenivasan KR. Direct determination of the f(a) singularity spectrum and its application to fully developed turbulence. Phys Rev A 1989;40:5284–94. [66] Schreiber T, Schmitz A. Improved surrogate data for nonlinearity tests. Phys Rev Lett 1996;77:635–8. [67] Zamir M. Critique of the test of multifractalty as applied to biological data. J Theor Biol 2003;225:407–12. [68] Montebelli A, Lowe R, Ziemke T. The cognitive body: from dynamic modulation to anticipation. In: Pezzulo G, Butz M, Sigaud O, Baldassarre G, editors. Anticipatory behavior in adapative learning systems. LNAI, vol. 5499. Springer; 2009. p. 132–51. [69] Holden JG, Van Orden GC, Turvey MT. Dispersion of response times reveals cognitive dynamics. Psychol Rev 2009;116:318–42. [70] Stephen DG, Dixon JA. The self-organization of insight: entropy and power laws in problem solving. J Problem Solving 2009;2:72–101. [71] Stephen DG, Boncoddo RA, Magnuson JS, Dixon JA. The dynamics of insight: mathematical discovery as a phase transition. Memory Cogn 2009;37:1132–49. [72] Stephen DG, Arzamarski R, Michaels CF. The role of fractality in perceptual learning: exploration in dynamic touch. J Exp Psychol Hum Percept Perform 2010;36:1161–73.

Multifractal cascade dynamics modulate scaling in ...

contrast between expectations of the temporal structure from the weak- and strong-anticipatory views. If intertap interval time series exhibited amax А amin not .... nents reported in [18] do not simply appear to have been. Table 1. Coefficients from the multiple linear regression model of amax А amin for intertap intervals.

441KB Sizes 2 Downloads 185 Views

Recommend Documents

Multifractal Dynamics in the Emergence of Cognitive ...
Harvard University, 3 Blackfan Circle, Floor 5, Boston, MA 02115. ... develop a formalism allowing us to understand how the broad spectrum of cognitive ... ena and the energy flow within the cognitive system (Stephen, Boncoddo, Magnuson, &.

Individual differences in mathematical competence modulate brain ...
e l s ev i e r. c om / l o c a t e / l i n d i f. Page 1 of 1. Individual differences in mathematical competence modulate brain responses to arithmetic errors.pdf.

Individual differences in mathematical competence modulate brain ...
Data from both neuropsychological and neuroimaging studies have ... task demands, such as working memory and attention, instead of .... Individual differences in mathematical competence modulate brain responses to arithmetic errors.pdf.

Scaling up predatorprey dynamics using spatial ...
explain population cycles, species coexistence, the effects of enrichment, or predict the evolution of behavioural traits, are often ... thanks to the development of adaptive dynamics theory ..... various phases of the cycle (Englund & Leonardsson, 2

Bodily Illusions Modulate Tactile Perception
se, that influences tactile processing. .... tion of touch might reflect a general process of cross- ..... parallel processing of tactual information in somatosensory.

practical-considerations-in-scaling-supercritical-carbon-dioxide ...
N. (1) Turbine specific speed calculation. Page 3 of 76. practical-considerations-in-scaling-supercritical-carbon-dioxide-closed-brayton-cycle-power-systems.pdf.

Cascade Middle School 4.10.pdf

Task Dynamics and Resource Dynamics in the ...
the Royal Society, London, Series B, 97, 155-167. Hinton, G. E. (1984). Parallel computations for controlling an arm. Journal of Motor Behavior, 16, 171-194.

Manual Scaling -
271. 280. 289. 298. 307. 316. 325. 334. 343. 352. Sine(Light). Cosine (Gravity). Tangent (Magnetism). MGL (Concentration). MG/L (Radiation). Manual Scaling.

Emotional stimuli modulate readiness for action: A ...
fields that cause neurons to depolarize (Bohning, 2000). If TMS is applied to .... mapping of congruency to response hand was counterbalanced across subjects.