Maintaining binding in working memory: comparing the effects of intentional goals and incidental affordances

Encode a Letter and Get Its Location for Free? Assessing Incidental Binding of Verbal and Spatial Features

Previous studies have demonstrated that when presented with a display of spatially arranged letters, participants seem to remember the letters’ locations when letters are the focus of a recognition test, but do not remember letters’ identity when locations are tested. This strong binding asymmetry suggests that encoding location may be obligatory when remembering letters, which requires explanation within theories of working memory. We report two studies in which participants focused either on remembering letters or locations for a short interval. At test, positive probes were either intact letter–location combinations or recombinations of an observed letter and another previously occupied location. Incidental binding is observed when intact probes are recognized more accurately or faster than recombined probes. Here, however, we observed no evidence of incidental binding of location to letter in either experiment, neither under conditions where participants focused on one feature exclusively for a block, nor where the to-be-remembered feature was revealed prior to encoding with a changing pre-cue, nor where the to-be-remembered feature was retro-cued and therefore unknown during encoding. Our results call into question the robustness of a strong, consistent binding asymmetry. They suggest that while incidental location-to-letter binding may sometimes occur, it is not obligatory.

Abnormal prefrontal and parietal activity linked to deficient active binding in working memory in schizophrenia

Working memory deficits have been widely reported in schizophrenia, and may result from inefficient binding processes. These processes, and their neural correlates, remain understudied in schizophrenia. Thus, we designed an FMRI study aimed at investigating the neural correlates of both passive and active binding in working memory in schizophrenia. Nineteen patients with schizophrenia and 23 matched controls were recruited to perform a working memory binding task, in which they were instructed to memorize three letters and three spatial locations. In the passive binding condition, letters and spatial locations were directly presented as bound. Conversely, in the active binding condition, words and spatial locations were presented as separated, and participants were instructed to intentionally create associations between them. Patients exhibited a similar performance to the controls for the passive binding condition, but a significantly lower performance for the active binding. FMRI analyses revealed that this active binding deficit was related to aberrant activity in the posterior parietal cortex and the ventrolateral prefrontal cortex. This study provides initial evidence of a specific deficit for actively binding information in schizophrenia, which is linked to dysfunctions in the neural networks underlying attention, manipulation of information, and encoding strategies. Together, our results suggest that all these dysfunctions may be targets for neuromodulation interventions known to improve cognitive deficits in schizophrenia.

Neither separate nor equivalent: relationships between feature representations within bound objects

Evidence suggests that binding, or encoding a feature with respect to other features in time and space, can convey cognitive advantages. However, evidence across many kinds of stimuli and paradigms presents a mixed picture, alternatively showing cognitive costs or cognitive advantages associated with maintaining bound representations. We examined memory for colored letters drawn from similar and distinct color sets under circumstances that encouraged or discouraged the maintenance of color-letter binding. Our results confirmed previous change recognition research showing feature recognition improvement under explicit instructions to maintain binding. Color memory improved during binding, showing a reduced detrimental effect of feature similarity on retrieval, particularly when the letter served as the retrieval cue for a letter-color object. We found that feature recognition improved when two conditions were met: 1) relationships between features were to-be-remembered, and 2) the feature conjunction was relevant at test. Our results further suggest that this feature advantage arises because the encoded relationship between the features facilitates retrieval, not because features and objects are represented simultaneously in separate buffers.

Visual short-term memory always requires general attention

The role of attention in visual memory remains controversial; while some evidence has suggested that memory for binding between features demands no more attention than does memory for the same features, other evidence has indicated cognitive costs or mnemonic benefits for explicitly attending to bindings. We attempted to reconcile these findings by examining how memory for binding, for features, and for features during binding is affected by a concurrent attention-demanding task. We demonstrated that performing a concurrent task impairs memory for as few as two visual objects, regardless of whether each object includes one or more features. We argue that this pattern of results reflects an essential role for domain-general attention in visual memory, regardless of the simplicity of the to-be-remembered stimuli. We then discuss the implications of these findings for theories of visual working memory.

Feature binding and attention in working memory: a resolution of previous contradictory findings

We aimed to resolve an apparent contradiction between previous experiments from different laboratories, using dual-task methodology to compare effects of a concurrent executive load on immediate recognition memory for colours or shapes of items or their colour-shape combinations. Results of two experiments confirmed previous evidence that an irrelevant attentional load interferes equally with memory for features and memory for feature bindings. Detailed analyses suggested that previous contradictory evidence arose from limitations in the way recognition memory was measured. The present findings are inconsistent with an earlier suggestion that feature binding takes place within a multimodal episodic buffer Baddeley, ( 2000 ) and support a subsequent account in which binding takes place automatically prior to information entering the episodic buffer Baddeley, Allen, & Hitch, ( 2011 ). Methodologically, the results suggest that different measures of recognition memory performance (A', d', corrected recognition) give a converging picture of main effects, but are less consistent in detecting interactions. We suggest that this limitation on the reliability of measuring recognition should be taken into account in future research so as to avoid problems of replication that turn out to be more apparent than real.