University of Pennsylvania Working

Papers in Linguistics
Volume 10
Article 4
Issue 2 Selected Papers from NWAVE 32

1-1-2005

Lenition of the Flap in American English

Matt Bauer

This paper is posted at ScholarlyCommons. http://repository.upenn.edu/pwpl/vol10/iss2/4

For more information, please contact [email protected].
Lenition of the Flap in American English

This working paper is available in University of Pennsylvania Working Papers in Linguistics: http://repository.upenn.edu/pwpl/
vol10/iss2/4
 1
Lenition of the Flap in American English

Matt Bauer

1 Introduction

This paper reports results from an experiment designed to test whether the
flap in American English exhibits a significant difference in the degree to
which it is lenited in certain domains. Results indicate the flap is weaker
within words compared to across words for males but not for females, who
produce flaps with consistently stronger articulations than males. The fact
that males exhibit variation in the quality of their flaps poses an interesting
problem for theoretical accounts that treat the flap as a categorical phenome-
non, but not a gradient one. This project represents a first step in determining
whether gradient aspects of the flap in American English are best understood
as a planned articulation resulting from a phonological process, a natural
outcome of phonetic implementation, or a combination of the two.

2 Background

The flap in American English is a lenited form of [t], [d], or [n] and it is trig-
gered when any of these segments occurs intervocalically before an un-
stressed syllable, as in water, ladder, and enter.
The domains under which flapping occurs are limited. Flapping does not
occur word-initially like in top, before a stressed syllable like in baton, or
after a fricative like in pasta or after (Banner-fnouye 1995). Flapping occurs
across word boundaries as in caught 'em, said it, and paid her, but not if the
boundary marks the terminus of an intonational contour, as in John had two
apples with his coffee, and Pete. eight bananas with his milk (Parker and
Walsh 1982).
Flaps are one of the shorter articulations in English, lasting about half as
long as a stop, and sometimes exhibiting only three to six pitch periods (Saw
1993, Ladefoged 1993). Despite a close relationship to taps and released
alveolar stops, flaps appear to represent a unique articulation. In an articula-
tory comparison of flaps to taps (as in Bertie and party), Saw (1993) found
that tongue-tip movements for flaps tend to be forward-backward, where for
taps, tongue movements are more up-down. The reason for the difference is

1
1 would like to thank Lisa Zsiga and Frank Parker for their comments on an
earlier version of this paper.

U. Penn Working Papers in Linguistics, Volume 10.2, 2005

 32 MATT BAUER

summarized by Sanner-Inouye (1995), who suggests the flap gesture, unlike

the tap and stop, is "ballistic" in nature, much like the hand of a pitcher re-
leasing a ball: the tongue tip is "cocked back" then "thrown" against the
alveolar ridge and quickly retracted. The uniqueness of flaps , however, IS
debated- this is discussed in the next section.

3 Theoretical Accounts of the Flap

Theoretical accounts of the flap generally fall into one of two categories: (l)
those that treat the flap as a unique consonant articulation and (2) those that
treat the flap as a byproduct of coarticulation. The first account, put forth by
Selkirk ( 1984), Kahn ( 1976), and to some extent by Sanner-Inouye ( 1995),
argues that flapping is the result of a planned change in articulation by
phonological environment- this is the standard, generative theory, which
models the change by a switch in the set of distinctive features that catego-
rize the consonant- thus this class of theories may be called "feature-based."
For example, Selkirk argues that the 't' in butter is characterized by the fea-
ture [+sonorant] (a member of the class of sounds with no constriction of
airflow), while the 't' in pasta is [-sonorant] (a member of a class of sounds
that do constrict airflow). Because the featural change is constant across the
different environments that cause flapping, feature-based theories predict
that any environment that produces a flap will produce only one kind of flap ,
and no significant measurable difference between flaps should exist. The
second account, offered by Browman and Goldstein (1990) and Browman
and Goldstein ( 1992), is a "gesture-based" theory, which proposes that flaps
are best understood as the result of an abstract instruction directing the
tongue-tip to achieve closure at the alveolar ridge. A benefit of this analysis
is that it claims to be able to explain without reference to rules and features
why certain sounds, despite a perceptual difference, are in fact the same.
The gestural account departs from feature-based theories by arguing that
flaps do not represent a uniquely planned articulation. Rather, flaps are es-
sentially released alveolar stops that have been "intruded" by surrounding
segments (de Jong, Beckman, and Edwards 1993; Browman and Goldstein
1992). The "open-closed-open acoustic contour" for intervocalic released
stops gives rise to the percept of a flap (butter); whereas, for released stops
after fricatives, the percept is a stop (after); thus, the difference between the
two is perceptual (Browman and Goldstein 1992). In a study comparing the
flap to alveolar stops, de Jong (1998) found that trained phoneticians listen-
ing to extractions of flaps and their surrounding vowels made a range of
judgments about the categorical status of flaps: from [d), to [t] to a flap. He
took this as evidence that the relationship between a flap and a full stop is
 LENfT!ON OF THE FLAP fN AMERICAN ENGLISH 33

scalar, not categorical. A limitation of that study is that the flaps that de Jong
used were all in the same phonological environment (across words: toad on,
or tote on), so it is not clear if the range of variation that the flaps showed
represented an unexpected patterning.
Several issues regarding the fl ap remain to be addressed. One question is
if flaps show predicable gradient properties such that some flaps are more
flap-like in some environments but more stop-like in other environments.
Feature-based theories predict that flaps are stab le articulations across any
environment that meets its structural description. Gesture-based theories
deny the uniqueness of flaps and allow for the possibility of gradient degrees
of closure. If two flapping environments produce two qualities of fl aps, then
feature-based acco unts will have been proven inadequate. However, if flaps
across minimally different environments are similar, then both feature- and
gesture-based theories model flapping equally well.

4 The Study

The study was designed to test whether speakers of American English ex-
hibit variation in the quality of their flaps in normal speech. Two factors
were chosen. The first factor was whether the quality of the flap would di ffer
across a morphological boundary compared to a word boundary. For exam-
ple, the italicized 't' in the underlined sections in (1-2) are both environ-
ments where a flap is expected to occur in normal speech. One question this
study addressed is whether the flap in (I) is produced qualitatively different
from the flap in (2).

(I) The examples were clearly stated in the paper.

(2) The authors wondered whether to state it in the introduction.

The second factor considered was whether the frequency with which a word
occurs in normal speech corresponds to a change in the quality of the flap .
For example, the verb note in the Kucera-Francis Word Frequency Corpus
occurs 486 times, with a log frequency of 398.9. In contrast, the verb allot
occurs only 13 times with a log frequency of 5.34. Both verbs end in [t], and
when either the pronoun it or the past tense marker -ed follows the segment,
it produces an environment where a flap is expected to occur. Several studies
have found frequency effects associated with the perception and production
of phonetic aspects of speech (Vitevich et al. 1999; Pierrehumbert 2000;
Hooper 1976), so it seems reasonable that word frequency may influence the
quality of the flap. So, a second question that this study addressed was
34 MATT BAUER

whether speakers produce flaps of differing quality in high-frequency words

like note compared to low-frequency words such as allot.

4.1 Participants

Five participants, 2 male and 3 female, volunteered to take part in the study.
The participants were selected because they were a friend or an acquaintance
of the researcher. All of the participants were native speakers of English and
originated from a variety of areas across the U.S .: Northern California, Ne-
braska, Seattle, Minneapolis, and rural Minnesota. The participants were not
paid.

4.2 Materials and Procedure

Ten high-frequency and ten low-frequency verbs ending in [t] were chosen
from the Kucera-Francis Word Frequency Corpus. The high-frequency
words had a log-frequency average of 155.24, and the low-frequency words
had an average of 7.15 (see appendix for complete list of verbs). Each verb
was embedded in two sentences (see appendix for list of sentences). In the
first, the verb was followed by past tense -ed, and in the second the pronoun
it. An additional 44 filler sentences were included.
Participants were given one of five sets of sentences. For each set, the
order in which the sentences were given was random, with the exception that
the first and last five were filler sentences. Also, participants were randomly
given one of two sentences that contained the same verb. For example, if a
participant was given sentence (3), she was not given sentence (4).

(3) The candidates debated on Tuesday.

(4) The candidates decided to debate it on Tuesday.

For any set of sentences, participants were given 10 high-frequency verbs

and I 0 low frequency verbs which included I 0 sentences with past tense -ed
and I 0 sentences with the pronoun it. Since measures were taken to ensure
participants were not given the same verb twice, they were given 20 target
sentences and 44 fillers in total.
Participants were given a packet of the sentences in 14-point Times New
Roman font. Sentences were separated by triple spacing. Each packet con-
sisted of six pages. Participants were asked to read each sentence silently
first, then read it aloud "in a normal way, at a normal loudness, and at a
normal rate of speaking." Participants were also told they had an opportu-
nity to rerecord a sentence if they stumbled on their words, coughed, or were
 LENfTrON OF THE FLAP IN AMERICAN ENGUSH 35

interrupted in any way. Three of 320 recorded sentences were rerecorded,

and only one was a target sentence.
A Telex M-40 microphone designed for desktop computers was used.
Each sentence was recorded and saved into individual files using the sound
recorder in Praat 4.0.! l running on a Hewlett-Packard 8660C Pentium Ill
PC.

4.3 Measurement and Analysis

Lenition is defined variously. Here, the use of " lenition" will follow Sanner-
Inouye ( 1995), who draws from Pagliuca and Mowrey ( l 987), in suggesting
that lenition is "the erosion of closure resulting from less radical muscular
movement." For the flap, a lenited flap is one where the tongue tip makes
less contact with the alveolar ridge than a normal flap.
In acoustic studies like this, contact must be measured indirectly. Flaps
are usually measured by their length in ms, and, as mentioned already, the
typical flap is about 30 ms and around three to six pitch periods. However,
duration is an inappropriate measure for comparing lenition among flaps
because the length of an articulation does not correlate with the degree of
contact between an articulator and its goal. Two flaps with an identical
length may nonetheless differ with respect to the degree to which they are
each lenited. For example, compare the flap in Figure I to the one in Figure
2 (marked by [DJ), produced by two separate speakers, M l and Ft. Both
flaps are approximately 38 ms, but the energy present in the wave of the flap
for M l compared to its surrounding vowels appears less than the energy pre-
sent for the wave of the flap compared to its surrounding vowels for Fl .
The energy present in the waves during the flap is key to observing leni-
tion defined here. The intuition is that energy, which is measured in decibels,
is related to the amount of contact an articulator makes with its goal : Deci-
bels directly measure, albeit on a logarithmic scale, the amplitude of sound
pressure changes, which is indirectly related to the degree of openness
(Johnson l 997). So, decreased energy of a flap compared to its surrounding
vowels indicates greater constriction of airflow, and more contact with the
alveolar ridge. The same flaps of M I and F I from Figures ( 1-2) graphed by
their intensities in decibels in Figures (3-4) show such a difference. Ml pro-
duces a flap with less of a difference in intensity between its two vowels than
Fl does. The deduction then is that Ml makes less contact with the alveolar
ridge in his flap than F l , and so, the flap of M l is more lenited than the flap
ofF I.
The measure of lenition in the experiment was obtained as follows . For
each of the target flapping environments, an extraction was taken beginning
36 MATT BAUER

with the onset of the preceding vowel (VI) of the flapping environment and
ending after the following vowel (V2) . Before an analysis was made, the
environment was judged by the researcher on whether the (t] was flapped or
not. The extracted sections were analyzed in Praat and graphed by their in-
tensity in dBs. For each graph, the point at which VI exhibited its greatest
intensity was averaged with the highest point of V2 to obtain a mean peak
vowel intensity. The lowest point on the graph corresponding most closely
with the location of the flap was subtracted from the mean vowel intensity to
obtain the distance in intensity that the flap differed from its surrounding
vowels. Strongly articulately flaps showed greater distance from the sur-
rounding vowels, and heavily lenited flaps showed lesser distance. Statistical
results were computed using SPSS 10.0.

ie D

0 0 .15
Time (s)
Figure 1: Waveform ofF! 's production of flap and surrounding vowels of
create it

0 0.15
Figure 2: Waveform of Ml 's production of flap and surrounding vowels of
limit it
 LENITION OF THE FLAP IN AMERICAN ENGLISH 37

65 .--------------------------------------------------.

[0
~
€
"'
c:
~

55+-----------------------~~----------------------~
0 0 .15
Figure 3: Intensity ofF I 's production of flap and surrounding vow-
els of create it
65 ~

m
~
~
.
c:
~

55+-------------------------------------------------~
0 0.15
Figure 4: Intensity of Ml 's production of flap and surrounding
vowels of limit it

5 Results

One hundred percent of the flapping environments exhibited a perceptible

flap.
Across all participants, a repeated-measures ANOVA showed that there
was no effect for either factor--for the morphological context, F(1 ,4)= 1.37,
p=.31, for the frequency context, F(l ,4)=4.78, p=.09--and no interaction,
F(l,4)= 1.48, p=.29 . However, individual ANOVAs performed separately on
each subject revealed that the two male participants significantly lenited the
flap within a morpheme context, "-ed," compared to across a word boundary,
"it," F( I, 19)=6. 78, p<.05 for the first male participant (M I), and
F( I, 19)= II. I 0, p<.05 for the second male participant (M2), but exhibited no
significant frequency effects, and no interaction. In contrast, the three female
participants (F I, F2, F3) did not exhibit any significant lenition as a result of
either factor. In fact, flaps produced by females had greater contrast with the
surrounding vowels compared to the flaps produced by males,
38 MATT BAUER

F(1 ,99)=25.47, p<.OOI. Results for all levels and factors are shown in Table
I.

MORPH*
MORPH FREQ FREQ
ALL 0.31 0.91 0.30
Ml **0.02 0.06 0.85
M2 **0.00 0.09 0. 10
Fl 0.99 0.76 0.99
F2 0.68 0.95 0.93
F3 0.71 0.64 0.78
Table I: Significance values from ANOV As, by morphological and fre-
quency factor

Graphs I and 2 show average difference in dB between the flap and its
surrounding vowels by gender and factor level.

Within word
fSl Across word

Male Femal e

Graph I: Relative difference in intensity between flap and its surrounding

vowels, by morphological location of the flap and gender of speaker, in dB
 - ------------------------------------------------------------------,

LENITION OF THE FLAP IN AMERICAN ENGLISH 39

Gil High frequ ency

E::l Low frequency

Male Female
Graph 2: Relative difference in intensity between flap and its surrounding
vowels, by word frequency and gender of speaker, in dB

It should be mentioned that that differences between male and female

values for the dependent variable is not a result of differences in physiology
(i.e. that females produce flaps between vowels with more of a difference in
energy than males by virtue of design). It is true that females tend to produce
higher-pitched speech than males, and so, one may expect higher-pitched
speech to exhibit more energy than lower-pitched speech (all other factors
being equal). But, the dependent variable used here measures relative decibel
difference between the flap and its surrounding vowels, not absolute levels
of the flap in isolation. Presumably, the measure is enough to abstract away
from differences between genders- and for that matter, differences between
speakers.
The difference in gender reported here is consistent with Byrd (1992),
cited in Byrd (1993), and Zue and Laferriera (1979). Byrd found that women
produce fewer segments judged as flaps than men, and Zue and Laferriera
found that men 's medial [t,d]s are significantly shorter than women's.
It is also worth noting that values for Ml and M2 by frequency factor
are close to significant. For these two males, flaps in high frequency words
appear to be articulated slightly stronger than flaps in low frequency words.
Interestingly, in a study of medial [t,d] deletion, Raymond, Dautricourt, and
Hume (2003) found that frequency does not predict whether the medial al-
40 MATT BAUER

veolar stop deletes. Clearly, more participants are needed to confirm or deny
the possibility that frequen cy plays a role in lenition of flappin g. Naturally,
this admission applies to gender differences as well.

6 Discussion

Results from participants M I and M2 demonstrate that the quality of the

flap is influenced by the morphological environment in which the flap oc-
curs. Across word boundaries, the flap is significantly lenited compared to
across morpheme boundaries. These results challenge the traditional view
that flaps do not exhibit varying qualities, as feature-based theories predict.
A solution for feature-based theories might be to alter the rule changing an
alveolar stop to a flap so that two flaps are permitted: one stronger than the
other. This seems plausible. The fact that I 00 percent of the [t]s in this study
were judged as flaps gives support for the conclusion that a phonological
rule alters the [t] to a full fl ap, and then there is further process that lenites
the flap within words. However, some evidence in the data reported here
suggests that such an alteration would still be inadequate: There appears to
be gradience between [t] and the flap as well. For example, even though all
the potential flaps that participant Fl produced were judged as flaps, a full
one-fourth of her " flaps" had detectible release bursts-a cue that these seg-
ments might actually be stops. This is in line with de Jong (1998) 's finding
that flaps occur along a continuum from a stop to a flap . In addition to gradi-
ence between a full stop and fl ap, several of the flaps produced by Ml and
M2 were nearly or completely deleted. Certainly, a single phonological rule
capturing strong, weak, and deleted flaps as well as flaps with release bursts
would hardly be able to support the conclusion that flaps represent a unique
consonant articulation independent from alveolar stops.
The alternative, gestural account of flapping seems appealing because it
can model scalar relationships between segments. Unfortunately, theories of
flapping under this model have either been untested or produced mixed re-
sults. For example, Browman and Goldstein's (1992) position that flapping is
the perception of a released stop occurring intervocalically remains to be
empirically confirmed. And de Jong's (1998) study of flaps across words
was unable to conclude that features or gestures alone could account for
flapping. This is not a criticism of de Jong's study, as it may tum out to be
the best way to understand flapping .
One obvious area for further research is a comparison of flaps and re-
leased alveolar stops within and across words. If the alveolar stops in words
like pasta and after and across words like cast her (krest#r] and lost her
[Iawst#r] resemble flaps with respect to (I) the distance in intensity between
 LENfTION OF THE FLAP IN AMERICAN ENGLISH 41

the stop and its surrounding vowels, and (2) the degree of lenition within and
across words , then that might suggest flaps and released stops differ only
perceptually. Further, if released alveolars show variation in its quality, from
flap to full stop, then the relationship among flap s and stops might, in fact,
be scalar.

Appendix

Word list
Low frequency High frequency
Word log freg Word log freg
debate 5.09 note 398.90
allot 5.34 treat 106.35
edit 5.38 complete 90.00
nominate 5.95 indicate 13.50
circulate 6.02 repeat 85.50
motivate 6.90 create 154. 10
donate 6.89 permit 138.76
recruit 8.57 operate 125. 10
bat 8.65 limit 121.23
spot 12.67 state 119.86

Sentence Jist
I . Attendance was limited to 500 guests.
2. Worried about violating fire codes, organizers decided to limit it to 500
guests.
3. The examples were clearly stated in the paper.
4. The authors wondered whether to state it in the introduction.
5. The construction worker operated the crane into the afternoon.
6. Everyone would operate it in the afternoon.
7. Mary wasn't permitted to speak.
8. The dog 's owner would never permit it to bark.
9. The company created a position for Janet.
I 0. The artist asked to create it for an exhibit.
II . Jim looked annoyed after he repeated his intentions.
12. Jim asked his wife to repeat it to Susan.
13 . Studies conducted in the past indicated a decline in interest.
14. Scientists almost never indicate it in their reports.
I 5. The students completed the survey in just 5 minutes.
16. Students were asked to complete it in 5 minutes.
17. The victim was treated for severe bums.
42 MATT BAUER

18. The victim wondered if the nurse would treat it with sa lve.
19. Several investigators noted the blood stain.
20. The investigators said they 'd note it in their summary.
21. The candidates debated on Tuesday.
22. The candidates decided to debate it on Tuesday.
23. The testing service allotted I 0 extra minutes for the verbal section thi s
year.
24. The teachers of third period refused to allot it for test-taking purposes.
25. The reporter usually edited hi s articles well ahead of deadline.
26. The reporter would usually edit it on Tuesday.
27. Six teachers at the high schoo l nominated the paper for an award.
28. Six teachers at the high school wanted to nominate it for an award.
29. The company donated several thousand dollars to the charity.
30. The company was happy to donate it to the charity.
3 I. The letter was circulated around the office.
32. Office workers encouraged Mark to circulate it around the office.
33. John 's dog was recruited for a dog-sledding race.
34. The dog-sledders hoped to recruit it for an upcoming race.
35. The baseball player batted at the ball.
36. The baseball player tried to bat it to left field.
37. Jane thought she spotted the car in the lot.
38. Jane looked for her car but couldn't spot it at all.
39. The cat wasn't particularly motivated to move for the guests.
40. The cat owners weren't successful to motivate it to move for the guests.

References

Banner-Inouye, Susan. 1995. Trills, taps, and stops in contrast and variation. Doc-
toral. Dissertation, UCLA.
Browman, Catherine. P. , and Louis Goldstein. 1990. Tiers in articulatory phonology,
with some implications for casual speech. In Papers in laboratory phonology 1:
Between the grammar and physics of speech, ed. J. Kingston and M. Beckman,
341-376. CUP.
-----. 1992. Articulatory phonology: An overview. Phonetica 49:155- 180.
Byrd, Dani. 1992. Sex, dialects, and reduction. In Proceedings of the International
Conference on Spoken Language Processing .
-----. 1993. 54,000 American Stops. UCLA Working Papers in Phonetics 83:97- I 16.
De Jong, Kenneth, Mary Beckman, and Jan Edwards. 1993. The interplay between
prosodic structure and coarticulation. Language and Speech 36: 197-212.
De Jong, Kenneth. 1998. Stress-related variation in the articulation of coda alveolar
stops: Flapping revisited. Journal of Phonetics 26:283-310.
Hooper, Joan. 1976. Word frequency in lexical diffusion and the source of morpho-
phonological change. In Current progress in historical linguistics, ed. W.
Christie, 96-105 Amsterdam: NorthHolland.
 LENlTlON OF THE FLAP lN AMERlCAN ENGLISH 43

Johnson, Keith . 1997. Acoustic and auditory phonetics. Blackwell.

Kahn , Daniel. 1976. Syllable-Based generalizations in English phonology. Doctoral
dissertation , MIT.
Ladefoged, Peter. 1993. A course in phonetics. Harcourt Brace Jovanovich.
Newman, R., J. Sawusch, & Paul Luce. 1997. Lexical neighborhood effects in pho-
netic processing. Journal of Experimental Psychology: Hum an Perception and
Performance 23(3):873 -889.
Pagliuca, W. and Mowrey R. (1987). Articulatory evolution. In Papersji-om the Vllth
Intern ational Conference on Historical Linguistics, ed. A. G. Ramal, 0 . Carruba
and G. Bernini, 459-472. Amsterdam: John Benjamins.
Pierrehumbert, Janet. 2000. Exemplar dynamics: Word frequency, lenition, and con-
trast. Ms. Northwestern University.
Parker, Frank & Thomas Walsh. 1982. Blocking alveolar flapping: A linguistic
analys is. Journal of Phonetics I 0:301-314.
Raymond, William; Robin Dautricourt, and Elizabeth Hume. 2003 . Word medial
/t,d/-deletion in spontaneo us speech: The effects of extra-linguistic,lexical,and
phonological factors. Ms. Ohio State University.
Saw, Cheng Cheng. 1993. Custom ized 3-D Electropalatography Display. UCLA
Working Papers in Phonetics 85:7 1-96.
Selkirk, Elizabeth. 1984. The syntax of words. Cambridge, Mass. : MIT Press.
Vitevitch, Mi chael S., Paul A. Luce, David B. Pisoni, and Edward T. Auer. 1999.
Phonotactics, neighborhood activation, and lexical access for spoken words.
Brain and Language 68( I ):306-311.
Zue, Victor, and Marta Laferriere. 1979. Acoustic study of medial /t,d/ in American
English. Journal of the Acoustical Society ofAmerica 66: I 039-1050.

Department of Linguistics
University of British Columbia
Vancouver, BC V6T IZI
matbauer@inlerchange. ubc. ca