Poster #6 - Measuring infant visual short term memory: An n-back approach

• In Event: 3-159 - Poster Session 12
  In Poster Session: PS 12 Section - Attention, Learning, Memory

Integrative Statement

Previous work demonstrates rapid improvements in infant visual short-term memory (VSTM) during the first year of life (e.g., Ross- Sheehy, Oakes, & Luck, 2003). Though findings based on looking time reveal clear improvements in VSTM performance, it is possible that eye movement metrics such as saccade and fixation counts may help identify the mechanisms involved in VSTM development. Using a modified n-back task, we will explore the extent to which serial fixation order interacts with infant characteristics such as saccade rate, and mean fixation duration to influence VSTM.

5- and 11-month-old infants viewed change-detection arrays consisting of four colored circles. Each trial began with a 1000ms sample array, followed by a 500ms retention interval, and finally a 3000ms test array, where one of the circles changed color. The location of the changed item was pseudorandomly chosen based on fixation order during the sample array and varied from 0 (the last item fixated) to -2. For example, if an infant fixated three items, the change location was equally probable to occur at either location 0 (the last item fixated), -1 (the second to last item fixated) or -2 (the third to last item fixated). Infants capable of remembering 1 item should detect the change at n-back = 0, whereas infants capable of remembering 2 items should detect the change at both n-back = 0 and -1. Note that due to low saccade counts during the sample array, only n-backs of 0 and -1 are reported.

Proportion of looking to the changed item during test was examined as a function of age (5 or 11mo) and saccade rate (low or high). Overall, infants looked significantly longer to the changed item for n-back of 0 than to an n-back of -1, F(1,42) = 31.702, p < .001. This main effect was qualified by an n-back by age by saccade rate three-way interaction, F(1,42) = 10.21, p = .003 (Fig1). Follow-up simple effect tests suggested that differences between high and low saccade groups were especially apparent at 5mos for n-back of 0, and at 11mos for an n-back of -1 (see Fig1). To further probe change detection across these distinct subject groups, we conducted a one-way ANOVA comparing proportion of looking to the changed item to chance. Results reveal significant change detection for n-back = 0 at 5mos for only the low saccade group, and at 11mo for the low saccade group, and marginally for the high saccade group (Fig1)

Taken together, these results suggest that infants with high saccade rates show relatively poor change detection, and this effect is particularly pronounced at 5 months. Additionally, these data suggest that change detection calculated based on proportion looking to the changed item is highly influenced by serial position effects. Though 11-month-olds do show some evidence memory for two items, this ability is apparent only in infants with relatively low saccade rates. These results suggest that individual differences in low-level visual behaviors such as saccade rate may contribute to emerging individual differences in cognitive development more generally.
 

Authors

• Bret Eschman, University of Tennessee, Knoxville
  Presenting Author

• Shannon Ross-Sheehy, University of Tennesee, Knoxville
  Non-Presenting Author