Christopher T. Kello

Scaling laws in cognitive sciences

Scaling laws are ubiquitous in nature, and they pervade neural, behavioral and linguistic activities. A scaling law suggests the existence of processes or patterns that are repeated across scales of analysis. Although the variables that express a scaling law can vary from one type of activity to the next, the recurrence of scaling laws across so many different systems has prompted a search for unifying principles. In biological systems, scaling laws can reflect adaptive processes of various types and are often linked to complex systems poised near critical points. The same is true for perception, memory, language and other cognitive phenomena. Findings of scaling laws in cognitive science are indicative of scaling invariance in cognitive mechanisms and multiplicative interactions among interdependent components of cognition.

The Emergent Coordination of Cognitive Function.

1/f scaling has been observed throughout human physiology and behavior, but its origins and meaning remain a matter of debate. Some argue that it is a byproduct of ongoing processes in the brain or body and therefore of limited relevance to psychological theory. Others argue that 1/f scaling reflects a fundamental aspect of all physiological and cognitive functions, namely, that they emerge in the balance of independent versus interdependent component activities. In 4 experiments, series of key-press responses were used to test between these 2 alternative explanations. The critical design feature was to take 2 measures of each key-press response: reaction time and key-contact duration. These measures resulted in 2 parallel series of intrinsic fluctuations for each series of key-press responses. Intrinsic fluctuations exhibited 1/f scaling in both reaction times and key-contact durations, yet the 2 measures were uncorrelated with each other and separately perturbable. These and other findings indicate that 1/f scaling is too pervasive to be idiosyncratic and of limited relevance. It is instead argued that 1/f scaling reflects the coordinative, metastable basis of cognitive function.

The Pervasiveness of 1/f Scaling in Speech Reflects the Metastable Basis of Cognition.

Human neural and behavioral activities have been reported to exhibit fractal dynamics known as 1/f noise, which is more aptly named 1/f scaling. Some argue that 1/f scaling is a general and pervasive property of the dynamical substrate from which cognitive functions are formed. Others argue that it is an idiosyncratic property of domain-specific processes. An experiment was conducted to investigate whether 1/f scaling pervades the intrinsic fluctuations of a spoken word. Ten participants each repeated the word bucket over 1,000 times, and fluctuations in acoustic measurements across repetitions generally followed the 1/f scaling relation, including numerous parallel yet distinct series of 1/f fluctuations. On the basis of work showing that 1/f scaling is a universal earmark of metastability, it is proposed that the observed pervasiveness of 1/f fluctuations in speech reflects the fact that cognitive functions are formed as metastable patterns of activity in brain, body, and environment.

Strategic control in word reading: evidence from speeded responding in the tempo-naming task.

To investigate strategic control over response initiation in word reading, the authors introduce the tempo-naming task. Relative to baseline performance in the standard-naming task, participants were induced to respond with faster latencies, shorter durations, and lower levels of accuracy by instructing them to time response initiation with an experimentally controlled tempo. The tempo response cue attenuated stimulus effects, and as faster tempos reduced latencies, the number of spelling-sound errors remained constant, whereas the number of word, nonword, and articulatory errors increased. To explain these results, the authors propose input gain as a mechanism of control over processing speed. The experimenters sketch how input gain could account for the current results as well as for the results from stimulus-blocking experiments testing the route emphasis and time criterion hypotheses of strategic control.

When Two Meanings Are Better Than One: Modeling the Ambiguity Advantage Using a Recurrent Distributed Network

Ambiguous words are processed more quickly than unambiguous words in a lexical decision task despite the fact that each sense of an ambiguous word is less frequent than the single sense of unambiguous words of equal frequency or familiarity. In this computer simulation study, we examined the effects of different assumptions of a fully recurrent connectionist model in accounting for this processing advantage for ambiguous words. We argue that the ambiguity advantage effect can be accounted for by distributed models if (a) the least mean square (LMS) error-correction algorithm rather than the Hebbian algorithm is used in training the network and (b) activation of the units representing the spelling rather than the meaning is used to index word recognition times.

Complexity matching in dyadic conversation

Recent studies of dyadic interaction have examined phenomena of synchronization, entrainment, alignment, and convergence. All these forms of behavioral matching have been hypothesized to play a supportive role in establishing coordination and common ground between interlocutors. In the present study, evidence is found for a new kind of coordination termed complexity matching. Temporal dynamics in conversational speech signals were analyzed through time series of acoustic onset events. Timing in periods of acoustic energy was found to exhibit behavioral matching that reflects complementary timing in turn-taking. In addition, acoustic onset times were found to exhibit power law clustering across a range of timescales, and these power law functions were found to exhibit complexity matching that is distinct from behavioral matching. Complexity matching is discussed in terms of interactive alignment and other theoretical principles that lead to new hypotheses about information exchange in dyadic conversation and interaction in general.

Strategic control over rate of processing in word reading: A computational investigation

It has recently been proposed that the time course of lexical processing can be influenced by strategic control factors in word reading. In the current study, a specific hypothesis of control, rate of processing, was implemented in three simulations of word reading. Simulation results showed that, by modulating the dynamic of processing, the control parameter input gain can account for effects of pressure for speed and stimulus blocking on naming performance. Results showed also that, to account for error patterns found in the tempo-naming task (Kello & Plaut, 2000), the influence of lexical knowledge must be strengthened relative to the standard dual-pathway architecture. Two methods of strengthening the influence of lexical knowledge are demonstrated: input gain used as a mechanism of route emphasis, or integration of the two commonly proposed pathways from orthography to phonology.

Situated Behavior and the Place of Measurement in Psychological Theory

Measured values of human behavior may entail contradictory attributes of wave and particle by analogy with the wave/particle attributes of the electron. 1/f scaling is the wave attribute in this analogy and punctate data points are the particle attribute. One consequence of the wave/particle duality in physics was to elevate measurement to a primary place in physical theory, and one purpose of the present analogy is to likewise elevate measurement to a primary place in psychological theory. Another purpose is to emulate Robert Shaws creative use of analogies, consistent with the brief quotation that begins this article.

Critical Branching Neural Networks.

It is now well-established that intrinsic variations in human neural and behavioral activity tend to exhibit scaling laws in their fluctuations and distributions. The meaning of these scaling laws is an ongoing matter of debate between isolable causes versus pervasive causes. A spiking neural network model is presented that self-tunes to critical branching and, in doing so, simulates observed scaling laws as pervasive to neural and behavioral activity. These scaling laws are related to neural and cognitive functions, in that critical branching is shown to yield spiking activity with maximal memory and encoding capacities when analyzed using reservoir computing techniques. The model is also shown to account for findings of pervasive 1/f scaling in speech and cued response behaviors that are difficult to explain by isolable causes. Issues and questions raised by the model and its results are discussed from the perspectives of physics, neuroscience, computer and information sciences, and psychological and cognitive sciences.

Large‐Scale Modeling of Wordform Learning and Representation

The forms of words as they appear in text and speech are central to theories and models of lexical processing. Nonetheless, current methods for simulating their learning and representation fail to approach the scale and heterogeneity of real wordform lexicons. A connectionist architecture termed the sequence encoder is used to learn nearly 75,000 wordform representations through exposure to strings of stress-marked phonemes or letters. First, the mechanisms and efficacy of the sequence encoder are demonstrated and shown to overcome problems with traditional slot-based codes. Then, two large-scale simulations are reported that learned to represent lexicons of either phonological or orthographic word-forms. In doing so, the models learned the statistics of their lexicons as shown by better processing of well-formed pseudowords as opposed to ill-formed (scrambled) pseudowords, and by accounting for variance in well-formedness ratings. It is discussed how the sequence encoder may be integrated into broader models of lexical processing.