Is awareness necessary for true inference?

Chained study and the discovery of relational structure

Prior knowledge of relational structure allows people to quickly make sense of and respond to new experiences. When awareness of such structure is not necessary to support learning, however, it is unclear when and why individuals "spontaneously discover" an underlying relational schema. The present study examines the determinants of such discovery in discrimination-based transitive inference (TI), whereby people learn about a hierarchy of interrelated premises and are tested on their ability to draw inferences that bridge studied relations. Experiencing "chained" sequences of overlapping premises during training was predicted to facilitate the discovery of relational structure. Among individuals without prior knowledge of the hierarchy, chaining improved relational learning and was most likely to result in explicit awareness of the underlying relations between items. Observation of chained training sequences was also more effective than the self-generation of training sequences. These findings add to growing evidence that the temporal dynamics of training, including successive presentation of overlapping associations, are key to understanding spontaneous relational discovery during learning.

Active transitive inference: When learner control facilitates integrative encoding

A growing body of research indicates that active control of learning improves episodic memory for material experienced during study. It is less clear how active learning impacts the integration of those experiences into flexible, generalizable knowledge. This study uses a novel active transitive inference task to investigate how people learn a relational hierarchy through active selection of premise pairs. Active control improved memory for studied premises as well as transitive inferences involving items that were never experienced together during study. Active learners also exhibited a systematic search preference, generating sequences of overlapping premises that may facilitate relational integration. Critically, however, advantages from active control were not universal: Only participants with higher working memory capacity benefited from the opportunity to select premise pairs during learning. These findings suggest that active control enhances integrative encoding of studied material, but only among individuals with sufficient cognitive resources.

Effects of spatial training on transitive inference performance in humans and rhesus monkeys

It is often suggested that transitive inference (TI; if A > B and B > C, then A > C) involves mentally representing overlapping pairs of stimuli in a spatial series. However, there is little direct evidence to unequivocally determine the role of spatial representation in TI. We tested whether humans and rhesus monkeys use spatial representations in TI by training them to organize 7 images in a vertical spatial array. Then, we presented subjects with a TI task using these same images. The implied TI order was either congruent or incongruent with the order of the trained spatial array. Humans in the congruent condition learned premise pairs more quickly, and were faster and more accurate in critical probe tests, suggesting that the spatial arrangement of images learned during spatial training influenced subsequent TI performance. Monkeys first trained in the congruent condition also showed higher test trial accuracy when the spatial and inferred orders were congruent. These results directly support the hypothesis that humans solve TI problems by spatial organization, and suggest that this cognitive mechanism for inference may have ancient evolutionary roots.

Detecting analogies unconsciously

Analogies may arise from the conscious detection of similarities between a present and a past situation. In this functional magnetic resonance imaging study, we tested whether young volunteers would detect analogies unconsciously between a current supraliminal (visible) and a past subliminal (invisible) situation. The subliminal encoding of the past situation precludes awareness of analogy detection in the current situation. First, participants encoded subliminal pairs of unrelated words in either one or nine encoding trials. Later, they judged the semantic fit of supraliminally presented new words that either retained a previously encoded semantic relation (“analog”) or not (“broken analog”). Words in analogs versus broken analogs were judged closer semantically, which indicates unconscious analogy detection. Hippocampal activity associated with subliminal encoding correlated with the behavioral measure of unconscious analogy detection. Analogs versus broken analogs were processed with reduced prefrontal but enhanced medial temporal activity. We conclude that analogous episodes can be detected even unconsciously drawing on the episodic memory network.

Integrating events across levels of consciousness

Our knowledge grows as we integrate events experienced at different points in time. We may or may not become aware of events, their integration, and their impact on our knowledge and decisions. But can we mentally integrate two events, if they are experienced at different time points and at different levels of consciousness? In this study, an event consisted of the presentation of two unrelated words. In the stream of events, half of events shared one component (“tree desk” … “desk fish”) to facilitate event integration. We manipulated the amount of time and trials that separated two corresponding events. The contents of one event were presented subliminally (invisible) and the contents of the corresponding overlapping event supraliminally (visible). Hence, event integration required the binding of contents between consciousness levels and between time points. At the final test of integration, participants judged whether two supraliminal test words (“tree fish”) fit together semantically or not. Unbeknown to participants, half of test words were episodically related through an overlap (“desk”; experimental condition) and half were not (control condition). Participants judged episodically related test words to be closer semantically than unrelated test words. This subjective decrease in the semantic distance between test words was both independent of whether the invisible event was encoded first or second in order and independent of the number of trials and the time that separated two corresponding events. Hence, conscious and unconscious memories were mentally integrated into a linked mnemonic representation.

Unconscious relational inference recruits the hippocampus

Relational inference denotes the capacity to encode, flexibly retrieve, and integrate multiple memories to combine past experiences to update knowledge and improve decision-making in new situations. Although relational inference is thought to depend on the hippocampus and consciousness, we now show in young, healthy men that it may occur outside consciousness but still recruits the hippocampus. In temporally distinct and unique subliminal episodes, we presented word pairs that either overlapped ("winter-red", "red-computer") or not. Effects of unconscious relational inference emerged in reaction times recorded during unconscious encoding and in the outcome of decisions made 1 min later at test, when participants judged the semantic relatedness of two supraliminal words. These words were either episodically related through a common word ("winter-computer" related through "red") or unrelated. Hippocampal activity increased during the unconscious encoding of overlapping versus nonoverlapping word pairs and during the unconscious retrieval of episodically related versus unrelated words. Furthermore, hippocampal activity during unconscious encoding predicted the outcome of decisions made at test. Hence, unconscious inference may influence decision-making in new situations.

The hippocampus reevaluated in unconscious learning and memory: at a tipping point?

Classic findings from the neuropsychological literature invariably indicated that performances on tests of memory that can be accomplished without conscious awareness were largely spared in amnesia, while those that required conscious retrieval (e.g., via recognition or recall) of information learned in the very same sessions was devastatingly impaired. Based on reports of such dissociations, it was proposed that one of the fundamental distinctions between memory systems is whether or not they support conscious access to remembered content. Only recently have we come to realize that the putative systemic division of labor between conscious and unconscious memory is not so clean. A primary goal of this review is to examine recent evidence that has been advanced against the view that the hippocampus is selectively critical for conscious memory. Along the way, consideration is given to criticisms that have been levied against these findings, potential explanations for differences in the reported results are proposed, and methodological pitfalls in investigations of unconscious memory are discussed. Ultimately, it is concluded that a tipping point has been reached, and that while conscious recollection depends critically on hippocampal integrity, the reach of the hippocampus extends to unconscious aspects of memory performance when relational memory processing and representation are required.

Mechanisms of inferential order judgments in humans (Homo sapiens) and rhesus monkeys (Macaca mulatta)

If A > B, and B > C, it follows logically that A > C. The process of reaching that conclusion is called transitive inference (TI). Several mechanisms have been offered to explain transitive performance. Scanning models claim that the list is scanned from the ends of the list inward until a match is found. Positional discrimination models claim that positional uncertainty accounts for accuracy and reaction time patterns. In Experiment 1, we trained rhesus monkeys (Macaca mulatta) and humans (Homo sapiens) on adjacent pairs (e.g., AB, BC, CD, DE, EF) and tested them with previously untrained nonadjacent pairs (e.g., BD). In Experiment 2, we trained a second list and tested with nonadjacent pairs selected between lists (e.g., B from List 1, D from List 2). We then introduced associative competition between adjacent items in Experiment 3 by training 2 items per position (e.g., B₁C₁, B₂C₂) before testing with untrained nonadjacent items. In all 3 experiments, humans and monkeys showed distance effects in which accuracy increased, and reaction time decreased, as the distance between items in each pair increased (e.g., BD vs. BE). In Experiment 4, we trained adjacent pairs with separate 9- and 5-item lists. We then tested with nonadjacent pairs selected between lists to determine whether list items were chosen according to their absolute position (e.g., D, 5-item list > E, 9-item list), or their relative position (e.g., D, 5-item list < E, 9-item list). Both monkeys' and humans' choices were most consistent with a relative positional organization.

Nonverbal transitive inference: Effects of task and awareness on human performance

We studied human nonverbal transitive inference in two paradigms: with choice stimuli orderable along a physical dimension and with non-orderable choice stimuli. We taught 96 participants to discriminate four overlapping pairs of choice stimuli: A+ B-, B+ C-, C+ D-, and D+ E-. Half of the participants were provided with post-choice visual feedback stimuli which were orderable by size; the other half were not provided with orderable feedback stimuli. In later testing, we presented novel choice pairs: BD, AC, AD, AE, BE, and CE. We found that transitive responding depended on task awareness for all participants. Additionally, participants given ordered feedback showed clearer task awareness and stronger transitive responding than did participants not given ordered feedback. Associative models (Wynne, 1995; Siemann and Delius, 1998) failed to predict the increase in transitive responding with increasing awareness. These results suggest that ordered and non-ordered transitive inference tasks support different patterns of performance.