The effect of non-communicative eye movements on joint attention

Nathan Caruana, Ayeh Alhasan, Kirilee Wagner, David M. Kaplan, Alexandra Woolgar, Genevieve McArthurQuarterly 2020 · DOI

TL;DR

This study reveals that while non-communicative eye movements can disrupt our ability to respond to joint attention cues, this effect is significantly less pronounced for human eye gaze compared to symbolic arrow cues. This suggests a specialized cognitive mechanism for processing social gaze, making us more resilient to distracting visual information when attending to communicative eye movements.

What Problem Does This Paper Address?

Joint attention, the ability to coordinate attention with another person to share an experience, is crucial for social development and interaction. Eye movements are key signals for initiating joint attention. However, in real-world social interactions, communicative eye gaze is often embedded within a stream of non-communicative eye movements (e.g., glancing away while listening). This raises a critical question: how do these "distracting" non-communicative eye movements influence our ability to respond to genuine joint attention bids? Previous research often used simplified scenarios with single, unambiguous gaze cues. This study aims to bridge this gap by investigating the impact of non-communicative eye movements, both predictable and unpredictable, on joint attention responsivity in a more dynamic, interactive setting, and comparing these effects to non-social visual cues.

Methods

The researchers employed an interactive computer game where participants, believing they were cooperating with another person ("Alan"), had to find a burglar. The core of the study involved manipulating the "search phase" before Alan initiated a joint attention bid. Three main "context conditions" were tested:

  • No Search: Alan made no distracting eye movements before initiating joint attention.
  • Random Search: Alan made a series of random, non-predictive eye movements before initiating joint attention.
  • Predictive Search: Alan's final non-communicative eye movement reliably predicted the location of the target.

Crucially, for each context condition, participants completed the task using either the avatar's eye gaze or dynamic arrow stimuli. This allowed for a direct comparison between social and non-social cue processing. Participants' saccadic reaction times (SRTs) to the joint attention cues were measured.

Datasets

Not discussed in the paper. The study used custom-designed experimental stimuli presented on a computer screen.

Key Insights and Ideas

  • Social Gaze Resilience: Our ability to respond to communicative eye gaze is remarkably stable, even when preceded by distracting, non-predictive eye movements. This "cost" of distraction is significantly lower for eyes than for arrows.
  • Arrow Vulnerability: Symbolic cues like arrows are much more susceptible to disruption. Unpredictable sequences of arrow movements severely impair response times, suggesting a less specialized processing mechanism for these cues.
  • The Power of Eye Contact: The study hints that initial eye contact (direct gaze) might play a role in preparing us for subsequent joint attention cues, potentially by modulating attention or arousal, making us more efficient in processing the intended signal.
  • Expertise in Social Gaze: Humans may have developed a specialized "expertise" in filtering out irrelevant information from social eye movements, a skill not as readily applied to symbolic cues like arrows.

Key Results

Accuracy: * In both experiments, participants made significantly more errors when non-communicative cues were non-predictive (Random Search) compared to when they were absent (No Search) or predictive (Predictive Search). This effect was observed for both eye gaze and arrow stimuli.

Saccadic Reaction Times (SRTs): * Experiment 1 (No Search vs. Random Search): * Participants were slower to respond to cues in the Random Search condition than in the No Search condition for both eyes and arrows. * The effect of context (Random vs. No Search) was significantly larger for arrows than for eyes. * In the No Search condition, participants were faster to respond to arrows than eyes. * Experiment 2 (Predictive Search vs. Random Search): * Participants were significantly slower to respond in the Random Search condition compared to the Predictive Search condition for both eyes and arrows. * A significant context by stimulus interaction was found, indicating a larger effect of context (predictability) for arrows than for eyes. * In the Predictive Search condition, participants were faster to respond to arrows than eyes.

Overall, the "cost" of non-predictive spatial information (Random Search) on response times was consistently smaller for eye gaze cues than for arrow cues across both experiments.

Limitations and Future Work

  • Ecological Validity of Arrow Stimuli: The authors acknowledge the challenge of creating perfectly matched social and non-social stimuli. While they made efforts to match arrows to eye movements, questions remain about whether the arrow stimuli accurately capture the "socialness" or lack thereof in natural interactions.
  • Role of Eye Contact Duration: The study suggests eye contact might be important, but future work is needed to systematically manipulate its presence and duration to confirm its role in facilitating joint attention.
  • Belief in Human Agency: The authors propose that manipulating participants' belief about whether the avatar/arrows are controlled by a human or computer could further elucidate the role of mentalizing mechanisms.
  • Developmental Trajectories: The findings highlight the need for future research to investigate how these contextual effects on gaze-following develop across different age groups.
  • Physiological Measures: The dynamic nature of the task presents methodological challenges, but future research could explore concurrent measures of arousal or attention (e.g., pupillometry) to better understand the underlying mechanisms.

Summary updated

Abstract

Eye movements provide important signals for joint attention. However, those eye movements that indicate bids for joint attention often occur among non-communicative eye movements. This study investigated the influence of these non-communicative eye movements on subsequent joint attention responsivity. Participants played an interactive game with an avatar which required both players to search for a visual target on a screen. The player who discovered the target used their eyes to initiate joint attention. We compared participants' saccadic reaction times (SRTs) to the avatar's joint attention bids when they were preceded by non-communicative eye movements that predicted the location of the target (Predictive Search), did not predict the location of the target (Random Search), and when there were no non-communicative eye gaze movements prior to joint attention (No Search). We also included a control condition in which participants completed the same task, but responded to a dynamic arrow stimulus instead of the avatar's eye movements. For both <i>eye</i> and <i>arrow</i> conditions, participants had slower SRTs in Random Search trials than No Search and Predictive Search trials. However, these effects were smaller for eyes than for arrows. These data suggest that joint attention responsivity for eyes is relatively stable to the presence and predictability of spatial information conveyed by non-communicative gaze. Contrastingly, random sequences of dynamic arrows had a much more disruptive impact on subsequent responsivity compared with predictive arrow sequences. This may reflect specialised social mechanisms and expertise for selectively responding to communicative eye gaze cues during dynamic interactions, which is likely facilitated by the integration of ostensive eye contact cues.
BibTeX
@inproceedings{caruana2020effect,
  author = {Nathan Caruana and Ayeh Alhasan and Kirilee Wagner and David M. Kaplan and Alexandra Woolgar and Genevieve McArthur},
  title = {{The effect of non-communicative eye movements on joint attention}},
  booktitle = {Quarterly},
  year = {2020},
  doi = {10.1177/1747021820945604},
}

Loading related content...

Figure