| Abstract: | Six pigeons were trained to asymptotic performance on a variable-delay matching-to-sample task in which the samples were sometimes line or color elements and sometimes line-color compounds. On compound-sample trials, the comparison stimuli were sometimes color elements and sometimes line-tilts. Sample type and delay (0, 1.5, and 4.5 sec) were varied within sessions, and sample duration (.4, 1.0, and 3.0 sec) was varied between sessions. Forgetting curves were steeper for line-tilt than for color. As sample duration increased, matching performance improved more for colors than for line-tilts, especially at delays greater than zero. Performance was better with element samples than with compound samples only on the line-tilt dimension at zero delay. Some predictions of a unitary trace growth and decay theory of pigeon short-term memory were not confirmed. A dual-code hypothesis was proposed to account for the data. |
