Nov 24
======

Clopper/Pisoni:2005
-------------------

Why did Peterson and Barney (1952) take only the first and second
formant frequency measurements?

In the work of Bradlow and Bent (2003), why did they report that
non-native listeners perform better than native listeners on speech
intelligibility tasks involving non-native speech samples? What type
of samples did they use?

Clopper and Pisoni recommend several ways to improve Preston's map
task, including exposing a listener to a short clip, requiring direct
perception of speech. Does this really get rid of any biases related
to working memory? Some variants are still more salient to naïve
listeners than others (perhaps that's the point) (pg. 316).

I understand that part of the reason Niedzielski (1999) (cited in
Clopper and Pisoni, 2005) used synthesized vowel tokens was to make
sure that the same "voice" was producing the speech and tokens (/aj,
aw/ as [ʌj, ʌw] before voiceless consonants) in question
(pg. 319). Have there been any studies that directly compared how
naïve listeners rate synthesized speech versus human talkers and the
accuracy of the dialect ratings? I think it would be difficult for one
speaker to try to talk with dialect features they don't possess (that
might sound 'unnatural' to a listener), so wouldn't it be better to
have different talkers produce speech? Although controlling the
talkers for other variables like gender and race/ethnicity raises some
issues as to who is chosen as the representative speaker of a certain
dialect (in the cases of the American studies, unless stated otherwise
they're white speakers).

How can one apply exemplar theory to the perception of linguistic
variation (besides experience as an effect, as suggested by Clopper
and Pisoni)? (p. 333)

Comments:
Even 20+ years after Purnll, Idsardi, and Baugh's (1999) study (cited
in Clopper and Pisoni, 2005), more recent research (Wright 2023) has
shown that 'non-standard' varieties of American English perform better
in neighborhoods with social demographics that match those of the
given non-standard variety. Not only does linguistic profiling occur
when hearing speech, but also reading it over text (Hanson and Hawley
2011, cited in Wright 2023). One interesting thing to note is that the
Fair Housing Act of 1968 protects homebuyers based on their membership
in protected classes (e.g. race, nationality, (dis)ability), but the
law is currently interpreted as one that must occur in physical
proximity (Hillier 2005, cited in Wright 2023). Therefore this
excludes interactions between potential homebuyers and landlords over
the phone or text.

Wright, K. (2023). Housing policy and linguistic profiling: An audit
study of three American dialects. Language Volume 99
(2). https://muse.jhu.edu/pub/24/article/900094/pdf

It was mentioned that a shift in methodology would allow additional
insights into the underlying process of how we as listeners make our
decisions in categorization tasks. One possible research question was
if some varieties might be easier to identify than others. But what
makes a variety easier to identify? Isn't that dependent on where the
listener and speaker is from?

It was stated that naïve listeners can make reliable judgements about
where an unfamiliar talker is from. However, the performance in
categorization tasks is pretty low (accuracy of around 30 %) even
after the perceptual training. Don't these results refute this
statement?

It is very interesting that participants rate their own language as
the most understandable and pleasant and that non-standard variants
are rated lower in terms of intelligence but higher in terms of
pleasantness.

It is also worth noting that our perception of speech sounds is
influenced by our knowledge and that this can be manipulated.

Why is the variation between white speakers much more regionally based
than variation among African-American speakers?

It is also interesting to note that women are generally ahead of men
in language change. Why is that?