Date Awarded


Document Type


Degree Name

Master of Science (M.Sc.)




John P Swaddle

Committee Member

Matthias Leu

Committee Member

Daniel Cristol


Collisions with windows are a leading anthropogenic cause of avian mortality. An estimated 500 million birds fatally collide with windows in North America every year. Visual mitigation strategies are thought to be an effective tool but there are very few robust field studies, and many controlled studies lack ecological relevance. This study seeks to test a window film in a more ecologically relevant manner, as well as introduce a novel element, sound. Previous work indicates that sound is an effective way to prevent collisions with freestanding structures. The sound is hypothesized to prompt a bird’s visual attention toward the collision hazard. Therefore, we further hypothesized that a sound cue (4-6 kHz oscillating frequency) used in concert with a visual cue (Solyx brand window film) would increase a bird’s ability to avoid a window. We exposed 24 zebra finches (Taeniopygia guttata) to four treatments (control, sound, visual, and sound-visual combined as a multimodal cue) in a balanced order in a flight chamber. A mist net ran the width of the flight corridor to prevent collisions from actually occurring. Each flight was filmed using multiple cameras. From these recordings, we reconstructed the flights in three dimensions and derived several key flight metrics (average velocity, end velocity, minimum distance to the structure, turn distance). We combined these metrics in a principal components analysis (PCA) that led to a quantitative categorization (using permutational multivariate analysis of variance, PERMANOVA) of “lower risk” (birds that made avoidance adjustments early in their flight) or “higher risk” (birds that made avoidance adjustments late in flight and those adjudged to collide with windows) flights. This categorization (lower vs higher risk) served as the response variable for a binomial generalized linear mixed model (GLMM). Based on the model, all three treatments (sound, visual, and multimodal) lowered the probability of riskier flight behavior, although the effect of the sound treatment alone was not as statistically supported as the other two treatments (95% confidence intervals overlapped with 0). Though the multimodal treatment appeared to have a slightly stronger effect than the visual alone, there is little statistical support that a multimodal approach is more effective than a unimodal (visual) one. We argue that multimodal strategies for preventing window collisions warrant further investigation and could have utility in situations where the efficacy of a visual treatment is compromised. While the effect of the sound cue alone was not strong, it was more effective than we predicted. Hence, sound may be an alternative deterrent in situations where applying a window film is not practical.



© The Author

Available for download on Monday, August 26, 2024

Included in

Biology Commons