Word learning in the field: Adapting a laboratory-based task for testing in remote Papua New Guinea

Abstract

Adapting laboratory psycholinguistic methods to fieldwork contexts can be fraught with difficulties. However, successful implementation of such methods in the field enhances our ability to learn the true extent and limitations of human behavior. This paper reports two attempts to run word learning experiments with the small community of Nungon speakers in Towet village in the Saruwaged Mountains, Papua New Guinea. A first attempt involved running a cross-situational task in which word-object pairings were presented ambiguously in each trial, and an explicit word learning task in which pairings were presented explicitly, or unambiguously, in each trial. While this quickly garnered a respectable 34 participants over the course of a week, it yielded null results, with many participants appearing to show simple patterned responses at test. We interpreted the null result as possibly reflecting the unfamiliarity of both the task and the laptop-based presentation mode. In Experiment 2, we made several adjustments to the explicit word learning task in an attempt to provide clearer instructions, reduce cognitive load, and frame the study within a real-world context. During a second 11-day stay in the village, 34 participants completed this modified task and demonstrated clear evidence of word learning. With this success serving as a future guide for researchers, our experiences show that it may require multiple attempts, even by experienced fieldworkers familiar with the target community, to successfully adapt experiments to a field setting.

Fig 1. Depiction of the Müller-Lyer illusion [4].

Although the two horizonal lines are the same length, the bottom line appears longer to those who perceive the illusion.

Fig 2. Examples of learning phase and test phase trials for Experiment 1.

Participants completed either the explicit word learning or XSWL training condition.

Fig 3. Proportion of correct responses (left) and reaction time (right) by participants who received training in the explicit learning condition or the ambiguous XSWL paradigm.

Error bars represent one standard error.

Fig 4. The left panel shows the randomized order of correct test responses.

The right panel represents participants’ pattern of key responses, with each row representing an individual’s responses. Both panels are standardized such that the first correct key (left) or response key (right) is represented by gray squares, with the alternate key represented by black squares. While correct responses were randomized, many participants appeared to show a patterning to their responses.

Fig 5. Example of the familiarization slide for Experiment 2.

Fig 6

Example of the instructional slides preceding the practice test (left two slides) and a walkthrough trial (right two slides) showing participants how to pick the correct referent (grass skirt) for the Nungon word bin.

Fig 7. The left panel shows the randomized order of correct test responses.

The right panel represents participants’ pattern of key responses, with each row representing an individual’s responses. Both panels are standardized such that the first correct key (left) or first response key (right) is represented by gray squares, with the alternate key represented by black squares.

Fig 8

Proportion of correct responses (left) and reaction time (right) at test by participants in the explicit word learning condition in Experiment 1 and Experiment 2. Error bars represent one standard error. **p < .01; ***p < .001.