Flowers Revisited: A Preliminary Replication of Flowers et al. 1997

TL;DR

This paper reports a preliminary replication of Flowers et al.'s 1997 cross-modal equivalence study, focusing on correlation estimation from visual, auditory, and audiovisual scatterplots. Using a within-subject design with 21 analyzed participants and 24 Gaussian datasets, the authors reproduce the original finding that perception of correlation is similar across modalities, with mean estimates closely tracking the true value (visual and audiovisual around 0.95, auditory around 0.85) and with greater auditory variability; audiovisual data showed visual dominance when both modalities were available. The work discusses replication culture in ICAD and proposes methodological refinements for a full-scale replication, including model selection via information criteria and separate analyses for positive and negative correlations, under the framing of communicative equivalence. Supplemental materials accompany the paper to support reproducibility. The study highlights the value of replication for establishing robust cross-modal perceptual guidance in sonification research.

Abstract

In 1997, Flowers, Buhman, and Turnage published a paper titled ``Cross-Modal Equivalence of Visual and Auditory Scatterplots for Exploring Bivariate Data Samples.'' This paper examined our capacity to assess the relationship between two data variables when presented through visual or auditory scatterplots. Twenty-seven years later, we have replicated the first part of this influential study and present the preliminary findings of our replication, initially involving 21 participants. In addition to purely auditory and visual scatterplots, we introduced audiovisual scatterplots as a third condition in our experiment. Our initial findings mirror those of Flowers et al.'s original research. With this extended abstract, we also aim to spark a discussion about the significance of replication studies for our research community in general.

Figures (4)

• Figure 1: Our participants' self-reported prior experience with sound and visualization, as well as their age and gender. For this preliminary reproduction, we did convenience sampling to recruit participants.
• Figure 2: Compare this figure to figure 2 from flowers_1997_CrossModalEquivalenceVisuala. The auditory estimations had the largest variability, resulting in a mean standard deviation of $13.34 mm$ (normalized to the $106 mm$ scale used by Flowers et al., who found a mean standard deviation of $10.84 mm$).
• Figure 3: Compare this figure to figure 3 from flowers_1997_CrossModalEquivalenceVisuala. For improved readability, we added dashed lines between the data points. The data appears more linear than the data collected in the original study. Nevertheless, the modalities perform similarly to each other, as they did in flowers_1997_CrossModalEquivalenceVisuala.
• Figure :