An associative account of the development of word learning

Word learning is a notoriously difficult induction problem because meaning is underdetermined by positive examples. How do children solve this problem? Some have argued that word learning is achieved by means of inference: young word learners rely on a number of assumptions that reduce the overall hypothesis space by favoring some meanings over others. However, these approaches have difficulty explaining how words are learned from conversations or text, without pointing or explicit instruction. In this research, we propose an associative mechanism that can account for such learning. In a series of experiments, 4-year-olds and adults were presented with sets of words that included a single nonsense word (e.g. dax). Some lists were taxonomic (i.,e., all items were members of a given category), some were associative (i.e., all items were associates of a given category, but not members), and some were mixed. Participants were asked to indicate whether the nonsense word was an animal or an artifact. Adults exhibited evidence of learning when lists consisted of either associatively or taxonomically related items. In contrast, children exhibited evidence of word learning only when lists consisted of associatively related items. These results present challenges to several extant models of word learning, and a new model based on the distinction between syntagmatic and paradigmatic associations is proposed.     

Statistical learning of adjacent and nonadjacent dependencies among nonlinguistic sounds

Previous work has demonstrated that adults are capable of learning patterned relationships among adjacent syllables or tones in continuous sequences but not among nonadjacent syllables. However, adults are capable of learning patterned relationships among nonadjacent elements (segments or tones) if those elements are perceptually similar. The present study significantly broadens the scope of this previous work by demonstrating that adults are capable of encoding the same types of structure among unfamiliar nonlinguistic and nonmusical elements but only after much more extensive exposure. We presented participants with continuous streams of nonlinguistic noises and tested their ability to recognize patterned relationships. Participants learned the patterns among noises within adjacent groups but not within nonadjacent groups unless a perceptual similarity cue was added. This result provides evidence both that statistical learning mechanisms empower adults to extract structure from nonlinguistic and nonmusical elements and that perceptual similarity eases constraints on nonadjacent pattern learning. Supplemental materials for this article can be downloaded from pbr.psychonomic-journals.org/content/supplemental.     

Implicit learning of second-, third-, and fourth-order adjacent and nonadjacent sequential dependencies

Serial reaction time (SRT) task studies have established that people can implicitly learn first- and second-order adjacent dependencies. Sequential confounds have made it impossible to draw conclusions regarding learning of nonadjacent dependencies and learning of third- and fourth-order adjacent dependencies. Addressing the confounds, the present study shows that people can implicitly learn second-, third-, and fourth-order adjacent and nonadjacent dependencies embedded in probabilistic sequences of target locations.     

Implicit learning of second-, third-, and fourth-order adjacent and nonadjacent sequential dependencies

Serial reaction time (SRT) task studies have established that people can implicitly learn first- and second-order adjacent dependencies. Sequential confounds have made it impossible to draw conclusions regarding learning of nonadjacent dependencies and learning of third- and fourth-order adjacent dependencies. Addressing the confounds, the present study shows that people can implicitly learn second-, third-, and fourth-order adjacent and nonadjacent dependencies embedded in probabilistic sequences of target locations.     

Perceptual processing of adjacent and nonadjacent tactile nontargets

Previous research has shown that subjects appear unable to restrict processing to a single finger and ignore a stimulus presented to an adjacent finger. Furthermore, the evidence suggests that, at least for moving stimuli, an adjacent nontarget is fully processed to the level of incipient response activation. The present study replicated and expanded upon these original findings. The results of Experiment 1 showed that an equally large response-competition effect occurred when the nontarget was presented to adjacent and nonadjacent fingers4on the same hand). The results of Experiment 2 showed that the effects observed in Experiment 1 (and in previous studies) were also obtained with stationary stimuli. Although small, there was some indication in the results of Experiment 2 that interference may dissipate more rapidly with distance with stationary stimuli. An additional finding was that interference effects were observed in both experiments with temporal separations between the target and nontarget of up to 100 msec. In Experiment 3, target and nontarget stimuli were presented to opposite hands. Although reduced, interference was still evident with target and nontarget stimuli presented to opposite hands. Varying the physical distance between hands did not produce any change in the amount of interference. The results suggest that the focus of attention on the skin extends nearly undiminished across the fingers of one hand and is not dependent upon the physical distance between sites of stimulation.     

Perceptual processing of adjacent and nonadjacent tactile nontargets.

Previous research has shown that subjects appear unable to restrict processing to a single finger and ignore a stimulus presented to an adjacent finger. Furthermore, the evidence suggests that, at least for moving stimuli, an adjacent nontarget is fully processed to the level of incipient response activation. The present study replicated and expanded upon these original findings. The results of Experiment 1 showed that an equally large response-competition effect occurred when the nontarget was presented to adjacent and nonadjacent fingers (on the same hand). The results of Experiment 2 showed that the effects observed in Experiment 1 (and in previous studies) were also obtained with stationary stimuli. Although small, there was some indication in the results of Experiment 2 that interference may dissipate more rapidly with distance with stationary stimuli. An additional finding was that interference effects were observed in both experiments with temporal separations between the target and nontarget of up to 100 msec. In Experiment 3, target and nontarget stimuli were presented to opposite hands. Although reduced, interference was still evident with target and nontarget stimuli presented to opposite hands. Varying the physical distance between hands did not produce any change in the amount of interference. The results suggest that the focus of attention on the skin extends nearly undiminished across the fingers of one hand and is not dependent upon the physical distance between sites of stimulation.     

Out of control: an associative account of congruency effects in sequence learning

The demonstration of a sequential congruency effect in sequence learning has been offered as evidence for control processes that act to inhibit automatic response tendencies (Jiménez, Lupiáñez, &; Vaquero, 2009) via unconscious conflict monitoring. Here we propose an alternative interpretation of this effect based on the associative learning of chains of sequenced contingencies. This account is supported by simulations with a Simple Recurrent Network, an associative (connectionist) model of sequence learning. We argue that the control- and associative-based accounts differ in their predictions concerning the magnitude of the sequential congruency effect across training. These predictions are tested by reanalysing data from a study by Shanks, Wilkinson, and Channon (2003). The results support the associative learning account which explains the sequential congruency effect without appealing to control processes (either conscious or unconscious).     

The indirect McCollough effect: An examination of an associative account

It has been found that viewing, for example, a red-and-black vertical grating alternating with a green homogeneous field produces a color aftereffect—a McCollough effect—on a black-and-white vertical grating (i.e., green). Viewing such colored patterns also produces an aftereffect on a noninduced horizontal grating (i.e., pink)—the indirect McCollough effect. Humphrey, Dodwell, and Emerson (1989) argued that the indirect McCollough effect is caused by opponent properties of the visual system that organize the processing of contour and color along contrasting, probably orthogonal, dimensions. Recently, however, their interpretation of the indirect McCollough effect has been challenged by some findings of Eissenberg, Allan, Siegel, and Petrov (1995). These researchers have proposed that the indirect McCollough effect, like the McCollough effect, can be explained by associative principles. The results reported here question crucial aspects of the hypothesis of Eissenberg et al.     

Contributions of the striatum to learning, motivation, and performance: an associative account

It has long been recognized that the striatum is composed of distinct functional sub-units that are part of multiple cortico-striatal-thalamic circuits. Contemporary research has focused on the contribution of striatal sub-regions to three main phenomena: learning of associations between stimuli, actions and rewards; selection between competing response alternatives; and motivational modulation of motor behavior. Recent proposals have argued for a functional division of the striatum along these lines, attributing, for example, learning to one region and performance to another. Here, we consider empirical data from human and animal studies, as well as theoretical notions from both the psychological and computational literatures, and conclude that striatal sub-regions instead differ most clearly in terms of the associations being encoded in each region.     

Testing an associative account of semantic satiation

How is the meaning of a word retrieved without interference from recently viewed words? The ROUSE theory of priming assumes a discounting process to reduce source confusion between subsequently presented words. As applied to semantic satiation, this theory predicted a loss of association between the lexical item and meaning. Four experiments tested this explanation in a speeded category-matching task. All experiments used lists of 20 trials that presented a cue word for 1 s followed by a target word. Randomly mixed across the list, 10 trials used cues drawn from the same category whereas the other 10 trials used cues from 10 other categories. In Experiments 1a and 1b, the cues were repeated category labels (FRUIT–APPLE) and responses gradually slowed for the repeated category. In Experiment 2, the cues were nonrepeated exemplars (PEAR–APPLE) and responses remained faster for the repeated category. In Experiment 3, the cues were repeated exemplars in a word matching task (APPLE–APPLE) and responses again remained faster for the repeated category.     

Filial imprinting and associative learning

Evidence is reviewed that is relevant to an interpretation of filial imprinting as a form of classical or Pavlovian conditioning. For such an interpretation it has to be assumed that the imprinting object provides both the unconditioned stimulus (UCS) and the conditioned stimulus (CS). The difficulty for an experimental approach is that CS and UCS are always presented simultaneously and cannot be manipulated independently, making it virtually impossible to test phenomena such as latent inhibition and extinction. Existing evidence is consistent with a conditioning theory of imprinting, for instance with regard to the reversibility of filial preferences. Some hitherto unexplained findings can be accounted for in terms of overshadowing and blocking. Results of an experiment are discussed in which the occurrence of blocking was explicitly tested. The findings support an associative learning interpretation of imprinting, but further experiments are needed to be able to rule cut alternative explanations.     

Exposure effects and associative learning

The contributions of initial stimulus affect and of associative learning to the effects of repeated stimulus exposures were examined in two experiments. Stimuli that were initially positive and stimuli that were initially negative were presented for different number of times, and subjects rated these stimuli afterward on a number of affective dimensions. In all cases, except when negative affect was associatively paired with every stimulus exposure, affective responses became increasingly more positive with increasing exposures. The results were taken to indicate that the exposure effect can overcome an initially negative stimulus affect when the conditions of the mere exposure hypothesis are satisfied. Initial stimulus affect and associative learning of affect were shown to be independent factors, the first influencing the intercept of the exposure function, the second its slope.     

An associative model of adaptive inference for learning word-referent mappings

People can learn word–referent pairs over a short series of individually ambiguous situations containing multiple words and referents (Yu & Smith, 2007, Cognition 106: 1558–1568). Cross-situational statistical learning relies on the repeated co-occurrence of words with their intended referents, but simple co-occurrence counts cannot explain the findings. Mutual exclusivity (ME: an assumption of one-to-one mappings) can reduce ambiguity by leveraging prior experience to restrict the number of word–referent pairings considered but can also block learning of non-one-to-one mappings. The present study first trained learners on one-to-one mappings with varying numbers of repetitions. In late training, a new set of word–referent pairs were introduced alongside pretrained pairs; each pretrained pair consistently appeared with a new pair. Results indicate that (1) learners quickly infer new pairs in late training on the basis of their knowledge of pretrained pairs, exhibiting ME; and (2) learners also adaptively relax the ME bias and learn two-to-two mappings involving both pretrained and new words and objects. We present an associative model that accounts for both results using competing familiarity and uncertainty biases.     

Sizing up the associative account of repetition priming

Three studies which test an associative account of repetition priming in a size comparison task are reported. Congruence of decision between priming and test affected performance when the priming task and test tasks were the same but not when they differed. This congruence effect was unaffected by the proportion of trials with congruent responses. Same-task priming exceeded cross-task priming even when both tasks required the same aspect of semantic knowledge. The results indicate that a component of priming is due to associations which are formed during priming and automatically activated when stimuli are repeated at test. Stimuli do not become associated with motor responses but are associated with the results of processing at a number of other levels.     

The development of a general associative learning account of skill acquisition in a relative arrival-time judgment task

Current theory assumes that individuals only use information from the immediate environment to perform relative arrival-time judgment tasks. This article presents a theoretical analysis of the memory requirements of this task. The authors present an analysis of the inputs to the memory system and the processes that map those inputs onto outputs. The analysis generates a set of predictions regarding the specificity of transfer and the role of context during learning. In 3 experiments, participants decided whether pairs of aircraft would violate a minimum separation standard or pass each other safely. Participants were presented with pairs of aircraft in which properties of the pair varied along 3 structural and 3 surface dimensions. Contexts were defined by the co-occurrence of specific values along stimulus dimensions and the use of a neutral label. The results suggest that transfer was limited by the dimensions that were varied in training and the context in which those dimensions were varied. The discussion focuses on the problems that complex tasks like relative judgment pose for associative learning mechanisms and the development of precise models of cognition.     

An associative model of rat serial pattern learning in three-element sequences

Hypotheses ranging from subsymbolic to symbolic have been proposed to account for rat sequential behaviour, and in the subsymbolic domain alone there are multiple proposed subsymbolic processes or factors thought to affect serial behaviour. A behavioural study and computer simulations were conducted to evaluate these hypotheses, and a new computational associative model based on pairwise associations and generalization was evaluated. Seven 3-element sequences were selected for study that systematically (1) varied sequence discriminability, (2) varied reward magnitude, and (3) manipulated the order of food quantities. Neither element discriminability nor response enhancement subsymbolic processes in isolation were able to account for the behavioural data; however, simulations from the computational model known as the sequential pairwise associative memory (SPAM) model with a log-linear mapping of stimulus dimension items to food quantities correlated well with the behavioural data. SPAM accounts for differential element anticipation in different sequences by appealing to pairwise association of sequence events and generalization between cues as the principal factors mediating pattern tracking in three-element sequences.     

An associative model of geometry learning: a modified choice rule

In a recent article, the authors (Miller & Shettleworth, 2007) showed how the apparently exceptional features of behavior in geometry learning ("reorientation") experiments can be modeled by assuming that geometric and other features at given locations in an arena are learned competitively as in the Rescorla-Wagner model and that the probability of visiting a location is proportional to the total associative strength of cues at that location relative to that of all relevant locations. Reinforced or unreinforced visits to locations drive changes in associative strengths. Dawson, Kelly, Spetch, and Dupuis (2008) have correctly pointed out that at parameter values outside the ranges the authors used to simulate a body of real experiments, our equation for choice probabilities can give impossible and/or wildly fluctuating results. Here, the authors show that a simple modification of the choice rule eliminates this problem while retaining the transparent way in which the model relates spatial choice to competitive associative learning of cue values.     

Distant melodies: statistical learning of nonadjacent dependencies in tone sequences

Human listeners can keep track of statistical regularities among temporally adjacent elements in both speech and musical streams. However, for speech streams, when statistical regularities occur among nonadjacent elements, only certain types of patterns are acquired. Here, using musical tone sequences, the authors investigate nonadjacent learning. When the elements were all similar in pitch range and timbre, learners acquired moderate regularities among adjacent tones but did not acquire highly consistent regularities among nonadjacent tones. However, when elements differed in pitch range or timbre, learners acquired statistical regularities among the similar, but temporally nonadjacent, elements. Finally, with a moderate grouping cue, both adjacent and nonadjacent statistics were learned, indicating that statistical learning is governed not only by temporal adjacency but also by Gestalt principles of similarity.     

Stimulus control and associative learning.

Interest in operant research on stimulus control has declined at the same time that much interest has burgeoned in nonoperant areas. Several examples of this shift toward traditional learning theory are considered, all of which have sponsored theoretical approaches that attempt to characterize the underlying associative units. These theoretical approaches are defended on the grounds that they have generated a deeper understanding of a variety of often puzzling phenomena. My projection is that future research will be determined even more strongly by theories about the structure of associations. Particular issues for which such discussion will have major impact include (1) whether conditional stimulus control is qualitatively different than simpler forms of stimulus control, (2) whether stimulus control is organized hierarchically, and (3) the origin of categories of stimulus equivalence.