Learning in noise: Dynamic decision-making in a variable environment

In engineering systems, noise is a curse, obscuring important signals and increasing the uncertainty associated with measurement. However, the negative effects of noise are not universal. In this paper, we examine how people learn sequential control strategies given different sources and amounts of feedback variability. In particular, we consider people’s behavior in a task where short- and long-term rewards are placed in conflict (i.e., the best option in the short-term is worst in the long-term). Consistent with a model based on reinforcement learning principles [Gureckis, T., & Love, B.C. Short term gains, long term pains: How cues about state aid learning in dynamic environments. Cognition (in press)], we find that learners differentially weight information predictive of the current task state. In particular, when cues that signal state are noisy, we find that participants’ ability to identify an optimal strategy is strongly impaired relative to equivalent amounts of noise that obscure the rewards/valuations of those states. In other situations, we find that noise and noise in reward signals may paradoxically improve performance by encouraging exploration. Our results demonstrate how experimentally-manipulated task variability can be used to test predictions about the mechanisms that learners engage in dynamic decision making tasks.     

Two pathways to stimulus encoding in category learning?

Category learning theorists tacitly assume that stimuli are encoded by a single pathway. Motivated by theories of object recognition, we evaluated a dual-pathway account of stimulus encoding. The part-based pathway establishes mappings between sensory input and symbols that encode discrete stimulus features, whereas the image-based pathway applies holistic templates to sensory input. Our experiments used rule-plus-exception structures, in which one exception item in each category violates a salient regularity and must be distinguished from other items. In Experiment 1, we found discrete representations to be crucial for recognition of exceptions following brief training. Experiments 2 and 3 involved multisession training regimens designed to encourage either part- or image-based encoding. We found that both pathways are able to support exception encoding, but have unique characteristics. We speculate that one advantage of the part-based pathway is the ability to generalize across domains, whereas the image-based pathway provides faster and more effortless recognition.     

Slowed Speech Input has a Differential Impact on On-line and Off-line Processing in Children’s Comprehension of Pronouns

The central question underlying this study revolves around how children process co-reference relationships—such as those evidenced by pronouns (him) and reflexives (himself)—and how a slowed rate of speech input may critically affect this process. Previous studies of child language processing have demonstrated that typical language developing (TLD) children as young as 4 years of age process co-reference relations in a manner similar to adults on-line. In contrast, off-line measures of pronoun comprehension suggest a developmental delay for pronouns (relative to reflexives). The present study examines dependency relations in TLD children (ages 5–13) and investigates how a slowed rate of speech input affects the unconscious (on-line) and conscious (off-line) parsing of these constructions. For the on-line investigations (using a cross-modal picture priming paradigm), results indicate that at a normal rate of speech TLD children demonstrate adult-like syntactic reflexes. At a slowed rate of speech the typical language developing children displayed a breakdown in automatic syntactic parsing (again, similar to the pattern seen in unimpaired adults). As demonstrated in the literature, our off-line investigations (sentence/picture matching task) revealed that these children performed much better on reflexives than on pronouns at a regular speech rate. However, at the slow speech rate, performance on pronouns was substantially improved, whereas performance on reflexives was not different than at the regular speech rate. We interpret these results in light of a distinction between fast automatic processes (relied upon for on-line processing in real time) and conscious reflective processes (relied upon for off-line processing), such that slowed speech input disrupts the former, yet improves the latter.     

Short-term gains, long-term pains: How cues about state aid learning in dynamic environments

Successful investors seeking returns, animals foraging for food, and pilots controlling aircraft all must take into account how their current decisions will impact their future standing. One challenge facing decision makers is that options that appear attractive in the short-term may not turn out best in the long run. In this paper, we explore human learning in a dynamic decision making task which places short- and long-term rewards in conflict. Our goal in these studies was to evaluate how people’s mental representation of a task affects their ability to discover an optimal decision strategy. We find that perceptual cues that readily align with the underlying state of the task environment help people overcome the impulsive appeal of short-term rewards. Our experimental manipulations, predictions, and analyses are motivated by current work in reinforcement learning which details how learners value delayed outcomes in sequential tasks and the importance that “state” identification plays in effective learning.     

There's more to emotion than meets the eye: A processing bias for neutral content in the domain of emotional prosody

Research on emotion processing in the visual modality suggests a processing advantage for emotionally salient stimuli, even at early sensory stages; however, results concerning the auditory correlates are inconsistent. We present two experiments that employed a gating paradigm to investigate emotional prosody. In Experiment 1, participants heard successively building segments of Jabberwocky "sentences" spoken with happy, angry, or neutral intonation. After each segment, participants indicated the emotion conveyed and rated their confidence in their decision. Participants in Experiment 2 also heard Jabberwocky "sentences" in successive increments, with half discriminating happy from neutral prosody, and half discriminating angry from neutral prosody. Participants in both experiments identified neutral prosody more rapidly and accurately than happy or angry prosody. Confidence ratings were greater for neutral sentences, and error patterns also indicated a bias for recognising neutral prosody. Taken together, results suggest that enhanced processing of emotional content may be constrained by stimulus modality.     

Navigating through abstract decision spaces: Evaluating the role of state generalization in a dynamic decision-making task

Research on dynamic decision-making tasks, in which the payoffs associated with each choice vary with participants’ recent choice history, shows that humans have difficulty making long-term optimal choices in the presence of attractive immediate rewards. However, a number of recent studies have shown that simple cues providing information about the underlying state of the task environment may facilitate optimal responding. In this study, we examined the mechanism by which this state knowledge influences choice behavior. We examined the possibility that participants use state information in conjunction with changing payoffs to extrapolate payoffs in future states. We found support for this hypothesis in an experiment in which generalizations based on this state information worked to the benefit or detriment of task performance, depending on the task’s payoff structure.     

The Processing of Non-canonically Ordered Constituents in Long Distance Dependencies by Pre-school Children: a Real-time Investigation

Four experiments were performed which had the goal of determining how and when young children acquire the ability to understand long distance dependencies. These studies examined the operations underlying the auditory processing of non-canonically ordered constituents in object-relative sentences. Children 4–6 years of age and an adult population participated in the study, which employed a cross modal picture priming methodology to determine when constituents in a non-canonical position are reactivated during ongoing sentence comprehension. The results support the view that even very young children have the same structural processing reflex seen in adults. Namely, children re-activate a non-canonically positioned (fronted) direct object NP immediately at the post-verb gap site during sentence processing.     

Speed of Lexical Activation in Nonfluent Broca''s Aphasia and Fluent Wernicke''s Aphasia

Rapid, automatic access to lexical/semantic knowledge is critical in supporting the tight temporal constraints of on-line sentence comprehension. Based on findings of “abnormal” lexical priming in nonfluent aphasics, the question of disrupted automatic lexical activation has been the focus of many recent efforts to understand their impaired sentence comprehension capabilities. The picture that emerges from this literature is, however, unclear. Nonfluent Broca's aphasic patients showinconsistent,notabsent,lexical priming, and there is little consensus about the conditions under which they do and do not prime. The most parsimonious explanation for the variable findings from priming studies to date is that the primary disturbance in Broca's lexical activation has something to do withspeedof activation. Broca's aphasic patients prime when sufficient time is allowed for activation to spread among associates. To examine this “slowed activation” hypothesis, the time course of lexical activation was examined using a list priming paradigm. Temporal delays between successive words ranged from 300 to 2100 msec. One nonfluent Broca's aphasic patient and one fluent Wernicke's patient were tested. Both patients displayed abnormal priming patterns, though of different sorts. In contrast to elderly subjects, who prime at relatively short interstimulus intervals (ISIs) beginning at 500 msec, the Broca's aphasic subject showed reliable automatic priming but only at a long ISI of 1500 msec. That is, this subject retained the ability to access lexical information automatically if allowed sufficient time to do so, a finding that may help explain disrupted comprehension of normally rapid conversational speech. The Wernicke's aphasic subject, in contrast, showed normally rapid initial activation but continued to show priming over an abnormally long range of delays, from 300 msec through 1100 msec. This protracted priming suggests failure to dampen activation and might explain the semantic confusion exhibited by fluent Wernicke's patients.     

Language performance on visual processing tasks in right hemisphere lesion cases

Seven language tests were constructed or adapted to assess the performance of three groups of 10 right-handed adult subjects: a right hemisphere lesion (RHL) group, a left hemisphere lesion (LHL) group, and a neurologically normal (NN) control group. Both the LHL and RHL groups produced poorer scores than the NN group on six of the seven tests. On two of the six significant tests, the RHL performed more poorly than the NN group. Analyses of words uttered during an oral story telling test indicated that the RHL group told significantly fewer complete stories using significantly more nouns, adjectives, and conjunctions than the NN group. On a 7-point scale, three judges rated the overall communication abilities of the RHL group as having “mild problems,” a significantly different rating than the ratings of the LHL and NN groups. The findings suggest that underlying visual spatial and perceptual deficits may be accompanied by clearly recognizable language differences in certain subjects.