Location coding with letters versus unfamiliar, familiar, and labeled letter-like forms.

Abstract

Abstract Errors in reporting a postcued target within a string of several letters more often involve a letter appearing elsewhere in the string (mislocation error) than one not appearing in the string (intrusion error). However, mislocation errors are not more frequent than intrusion errors with strings of relatively unfamiliar symbols or forms, and two experiments were conducted to determine the reason for this difference. The first experiment showed that the difference is not an artifact of different overall accuracy levels or exposure durations for the two character types. The second experiment revealed that, relative to no experience, initial experience with the forms through familiarization or learning labels for them produced mislocation and intrusion frequencies that were more similar to those for letters. This change was related to the development of stored representations for forms that are similar to those for letters and are subject to the same influences that produce mislocation and intrusion patterns for them.Determining relative positions may be more difficult for some types of items than for others. Chastain (1994) presented displays comprised of a row of only letters or only symbols (slashes, parentheses, etc.). One character was cued for report by an indicator presented in the postexposure pattern mask, and the subject responded by pressing a key on a computer keyboard corresponding to that character. Subjects could make two types of errors: (1) indicating an item that was present elsewhere in the display but appeared to be in the postcued position (called a mislocation error), or (2) indicating an item that was not present anywhere in the display (an intrusion error). Mislocation errors were more frequent than intrusion errors for displays containing only letters, but not for displays containing only symbols. More mislocations than intrusions have been observed by others when letters were the stimuli (Butler & Currie, 1986; Butler, Mewhort, & Tramer, 1987; Chastain, 1990). The nature of the mental representations of stimuli producing these results is the focus of the current study.Mewhort and Campbell (1981) proposed a Dual-Buffer model involving both letter identity and location. Information is transferred from an initial spatially faithful precategorical representation (Feature Buffer) to a postcategorical representation (Character Buffer) containing items with identity codes and relative position tags. Separate processing of identity and location information during the transfer from the Feature Buffer to the Character Buffer allows scrambling that produces mislocations (Butler, 1981). Others have proposed that location information merely decays more quickly than identity information, leaving observers less sure of the relative positions of items than of their identities (Dixon, 1986; Dixon & Twilley, 1988). More recently, Driver and Baylis (1991) proposed that letter identities are derived in parallel rather than serially, although letter location information is not automatically produced when letter identity is determined. Encoding location for letters may require that attention be directed to them serially.A study by Mewhort, Campbell, Marchetti, and Campbell (1981) confirmed the proposition that letter location information is fragile. They presented a postexposure mask containing a probe indicating the letter to be reported. Intrusion errors decreased markedly and in a regular fashion with increases in the delay between the offset of the target and the onset of the mask with probe. However, location errors decreased only slightly and erratically. Other research has shown that masking increases location errors far more than intrusion errors (Butler, Mewhort, & Tramer, 1987; Mewhort, Huntley, & Duff-Fraser, 1993).Letters are familiar stimuli with phonemic or name codes that are readily available. Perhaps identity information exists in the Character Buffer in the form of such codes, with location information in the form of a location tag associated with each code. …