Bias in the perception of phonetic detail in children’s speech: A comparison of categorical and continuous rating scales

Abstract

Previous research has shown that continuous rating scales can be used to assess phonetic detail in children’s productions, and could potentially be used to detect covert contrasts. Two experiments examined whether continuous rating scales have the additional benefit of being less susceptible to task-related biasing than categorical phonetic transcriptions. In both experiments, judgments of children’s productions of /s/ and /θ/ were interleaved with two types of rating tasks designed to induce bias: continuous judgments of a parameter whose variation is itself relatively more continuous (gender typicality of their speech) in one biasing condition, and categorical judgments of a parameter that is relatively less-continuous (the vowel they produced) in the other biasing condition. One experiment elicited continuous judgments of /s/ and /θ/ productions, while the other elicited categorical judgments. The results of Experiment 1 showed that the influence of acoustic characteristics on continuous judgments of /s/ and /θ/ was stable across biasing conditions. In contrast, the results of Experiment 2 showed that the influence of acoustic characteristics on categorical judgments of /s/ and /θ/ differed systematically across biasing conditions. These results suggest that continuous judgments are psychometrically superior to categorical judgments, as they are more resistant to task-related bias.

What accounts for individual differences in susceptibility to the McGurk effect?

“…The voiced consonant (/dɑ/) always appeared on the left end of the line, and the unvoiced consonant (/tɑ/) always appeared on the right end of the line. To be consistent with prior research, the VAS line was unlabeled, but the values ranged from 0 (/dɑ/) to 535 (/tɑ/), with a midpoint of 268 [ 76 , 78 ]. Participants were encouraged to use the entire line if they felt it was appropriate [ 75 , 76 , 78 ].…”

“…Perceptual gradiency was measured using a continuous VAS task, which has been shown to be sensitive to individual differences in phoneme categorization [ 53 , 73 – 78 ], and is less susceptible to task-related biases than categorical judgments [ 76 ]. In VAS tasks, participants are provided with a line with endpoints representing the extremes of a continuum, like /s/ on the left end and /∫/ on the right end of a centroid frequency continuum [ 73 ].…”

“…Stimuli for the VAS task consisted of a /dɑ/ to /tɑ/ continuum varying in both voice onset time (VOT) and fundamental frequency (f 0 ). We considered using several different continua, but since much of the existing research using the VAS task to measure perceptual gradiency relies on a single continuum [ 53 , 76 , 78 ], we opted to only use the /dɑ/ to /tɑ/ continuum. Stimuli were obtained from Kong and Edwards [ 53 ], and consisted of six log-scale VOT steps, and at each VOT step there were five f 0 steps (for more information about stimulus creation, see [ 53 ]).…”

Evidence-Based Survey Design: The Use of Continuous Rating Scales in Surveys

“…With the benefits reviewed in these studies, continuous rating scales such as VASs and sliders are frequently used in healthcare and the social sciences—for example, in tinnitus, endometriosis pain treatment, and orthopedics studies (Adamchic et al., ; Bourdel et al., ; Wall et al., ); in measuring happiness and patient satisfaction (Studer, ; Voutilainen et al., ); and in linguistics (Munson, Schellinger, & Edwards, ). Researchers in these studies found paper‐ or web‐based VASs or sliders to be reliable measurement instruments, and recommended them over 5‐, 7‐, or 10‐point discrete rating scales.…”

Individual Differences in Categorical Judgment of L2 Stops: A Link to Proficiency and Acoustic Cue-Weighting

“…Our study is inspired by two existing studies on the L1 perception that looked into English-speaking adults’ utilizations of multiple acoustic cues (primary cue of VOT and secondary cue of F0, that is, fundamental frequency at the post-consonantal vocalic onset for voiced-voiceless stops contrast) and their influence on the gradient manner of category judgments (voiced vs. voiceless stops; Kapnoula et al., 2017; Kong & Edwards, 2016). Instead of adopting the traditional forced-choice identification task, these studies administered a visual analog scaling (VAS) task that asked listeners to judge stimulus sounds not in a binary way (i.e., one category or the other) but in a continuous way (see Figure 1(a)) (Massaro & Cohen, 1983; Munson et al., 2017; Schellinger et al., 2017). Since listeners are allowed to click any locations along the arrow depending on the relative closeness of the stimuli to two given categories, this task encourages listeners to judge the category membership of the stimuli on a gradient scale and makes it possible to quantify individual listeners’ dichotomic or gradient decisions of speech categories.…”