Professional Documents
Culture Documents
Contextual Variability in Am Eng Dark L
Contextual Variability in Am Eng Dark L
Abstract
This paper presents a four-subject study that examines the relative influence of syllable position and
stress, together with vowel context on the colouring of the dark-l characteristic of speakers of General
American English. Most investigators report lighter /l/ tokens in syllable onsets and darker tokens in
coda positions. The present study demonstrates that when dark-l serves as an onset in iambic
intervocalic context with tautosyllabic high front vowels, it is fully dark as a result of domain-initial
strengthening. By contrast, when dark-l is abutted across a word boundary to word-final or word-
initial consonants, or when it is contained in a foot-internal context (preboundary intervocalic rime
with trochaic stress) its dorsal gesture is constrained, resulting in less dark tokens. In the case of dark-
l, articulatory undershoot must be understood not only in terms of the alveolar gesture, but also the
dorsal gesture.
Introduction
The phoneme /l/ has been the subject of increased investigation in recent years, in part due
to its varied phonetic characteristics under the influence of prosodic and metrical forces,
together with articulatory constraints of the short- and long-range phonetic environments.
The range of research tools used has increased over the years allowing for richer description
of its production, and its behaviour under contextual variation has been used to support a
range of theoretical constructs.
Articulation of /l/
The English /l/ is the alveolar lateral approximant. It has two main variants light-l (or clear-
l) and dark-l. Both variants are usually voiced, have tongue tip contact with the alveolar
ridge, and have one or both sides of the tongue lowered to allow lateral escape of air.
Sounds that follow the /l/ influence lip shape, and there may be a preference for lip
rounding for the dark-l by some speakers. The presence of a secondary dorsal gesture
Correspondence: Judith Oxley, Department of Communicative Disorders, University of Louisiana at Lafayette, PO Box 43170,
Lafayette, LA 70504-3170, USA. Tel: +1 337 482 1188. Fax: +1 504 482 6195. E-mail: odj8147@louisiana.edu
ISSN 0269-9206 print/ISSN 1464-5076 online # 2007 Informa UK Ltd
DOI: 10.1080/02699200701356485
524 J. Oxley et al.
differentiates dark-l from clear-l. The physiological literature suggests that the secondary
gesture may be of several types: upward movement toward the velum (velarizing),
backward movement (retraction), and pharyngeal retraction and lowering (Narayanan,
Alwan, & Huker, 1997; Sproat & Fujimura, 1993; Giles & Moll, 1975; Gick, Kang, &
Whalen, 2002). Thus, there appears to be considerable variation in the production
mechanisms of this secondary articulatory gesture. Recasens (2004) claims that darkness in
/l/ is scalar phonetic property; specifically, he claims that it may be more accurate to regard
the secondary gesture as essential to /l/, and that its involvement is a matter of degree.
There is increasing support for this position from a convergence of physiological evidence
from several methodologies. Acoustic data have also demonstrated that systematic changes
in boundary strengths, together with other factors clearly lead to corresponding lightening
or darkening of /l/ relative to the two extremes within a language or dialect. Crosslinguistic
and dialect studies reveal that both dark-l and light-l vary greatly even within languages or
dialects that are characterized by just one variant of /l/ (Recasens, 1995; Carter, 2003). In
the following sections we shall review three sets of variables that account for the varieties of
/l/ that occur in English (and evidently other languages): dialect, speaker characteristics
(‘‘idiolect’’), and phonetic factors.
Dialect
The absolute quality of /l/ colouring and its distribution seem to be dictated first by the
speaker’s dialect. It is well known that both clear-l and dark-l may co-vary within a
language: in Received Pronunciation English (RP) /l/ is distributed with clear-l typically in
onset position, and dark-l in coda position (Bladon & Al Bamerni, 1976; Bladon & Nolan,
1977); but, by contrast, one form may be used exclusively in both positions, such as clear-l
in dialects of English spoken in Northern Ireland, or dark-l in some Scottish dialects
(Carter, 2003). Tokens of /l/ in General American English (GAE), while generally
following the pattern of relatively clear onsets and darker codas seen in RP, are frequently
described in the literature as being darker overall than their RP counterparts, and particular
dialects of American English, especially Southern, are reported to include either
particularly dark varieties of /l/, or possible even the velar lateral, /L/ (Wells, 1982).
Recasens (2004) notes that American dark-l is darker than the Catalan and Majorcan dark-
l, which have low F2 values. These differences in distribution of clear-l and dark-l within
English dialects, however, do not create phonemic contrasts. In some languages clear-l and
dark-l are contrastive (e.g. Russian). Just as some English dialects are now non-rhotic, there
is evidence that particular English dialects (e.g. varieties of British and Australian English)
are characterized by consistent vocalization of coda-l (Hardcastle & Barry, 1989; Johnson &
Britain, 2003). Within GAE, however, it is possible that instances of vocalized /l/ (or
something very like vocalized-l) could sometimes represent genuine allophonic variations of
/l/ itself, or at other times reflect articulatory undershoot of the alveolar gesture.
Articulatory undershoot may be due to factors such as more rapid speech rate or influence
of surrounding consonants, such as /f/ (Giles & Moll, 1975). A clinical perspective is that l-
vocalization constitutes a developmental error, and that its persistence reflects a deviant
production (e.g. Bauman-Waengler, 2004). Patterns of /l/ usage, including deletions and
substitutions, strongly identify a speaker with a dialect. English dialects differ significantly
in terms of vowel distribution. For example, American English dialects differ greatly in
terms of range and distribution of back vowels and diphthongs (Ball & Müller, 2005; Roca
& Johnson, 1999; Wells, 1982). Given the strong evidence presented by Gick et al. (2002)
Contextual variability in American English dark-l 525
for coarticulation of the /l/ with the back vowel / /, it is likely that /l/ colouring will also be
influenced by the speaker’s vowel system. The degree of lip rounding that accompanies
dark-l will lower all formant values and further ‘‘darken’’ the resulting token. An excellent
review of the continuum of darkness of /l/ is available in Recasens and Espinosa (2005).
most common contexts for vocalization (vowelization) include when /l/ follows a back
vowel or forms a final cluster (e.g. in the words golf or, in some dialects built). The dorsal
gesture for dark-l is almost identical to that found in the vowel / / (Gick et al., 2002).
temporal and articulatory constraints to the maximum dorsal gesture, and these constraints
prevent darkening of /l/. From a strict metrical perspective, the boundary created in the
iambic context is one between two feet, and is stronger than the within-foot boundary
created in the trochaic counterpart (`deal#a). The front vowel trochaic context (deal a)
actually produced /l/s of comparable quality to the canonical onset position. Gestural
separation and timing, together explain this finding. The dorsal gestures of /i/ and dark-l are
in opposition, and so /l/ seems to affiliate closely with the following schwa, which shares the
Dorsal +Back feature. In addition, the foot-internal trochaic intervocalic context CV#c
interval was significantly shorter in duration than its iambic counterpart, creating a context
for articulatory undershoot of the dorsal gesture.
Clinical implications
From a clinical perspective, changes in the production of /l/ may result in errors: when
people produce onset-l as the glide /w/, or something very like it, the listener is apt to
perceive either a different word, or to judge the speech as being immature. The
phonological process identified as gliding is one of the last to be suppressed, and so children
may continue to say ‘‘wabbits’’ for ‘‘rabbits’’ and ‘‘weaves’’ for ‘‘leaves’’ in the earliest years
of formal schooling (Grunwell, 1987). Persistence of this pattern into middle childhood
years would be considered inappropriate. Therapy models typically facilitate the liquid/
glide contrast by moving the child from a stage of single-word through connected speech
production. The target sound is typically presented in a word with an initial /l/ (onset), and
then medially (intervocalic contexts, such as believe or Sally), and eventually embedded
within a longer phrase or sentence. Some treatment programmes emphasize the importance
of the phonetic environment as a variable for the clinician to control to prevent unhelpful
patterns of coarticulation (e.g. Shine, 1989; Bauman-Waengler, 2004). A glance through
samples of common articulation/phonology remediation materials marketed to clinicians
reveals that there is generally little attempt to differentiate ‘‘l-words’’ except for the most
general ‘‘initial-medial-final’’ distinctions (e.g. Artic Photos Fun DecksH For Articulation
& Language, Webber, 2006). The medial context arises within and across words, and
permits onsets, and codas that are actually ambisyllabic. Prosody can determine the quality
of consonants: for example, /t/ in a trochaic intervocalic position is typically realized as a
flap. In the case of dialects with onset dark-l, such as many variants of American English,
one would predict that any phonetic context that permitted a very dark /l/ would be
particularly subject to gliding by individuals predisposed to this pattern.
Purpose of study
In this paper, we shall investigate /l/ in a variety of phonetic and prosodic contexts with a
view to determining which acoustic properties are salient. This information could be
helpful to clinicians who are grappling with the problem of trying to facilitate their clients’
acquisition of the liquid/glide contrast.
Our hypotheses were derived from findings in existing studies that evaluated /l/ in either
trochaic or iambic contexts, but not both within the same speaker (Huffman, 1997; Sproat
& Fujimura, 1993). In these studies, the expectation was for a pattern of /l/ distribution
with light-l onset and dark-l coda; however, based on our earlier findings, we developed a
second set of hypotheses to address the possibility of /l/ distribution with dark-l in both
onset and coda positions.
528 J. Oxley et al.
Method
Participants
The four male speakers in this study were not randomly selected. They were chosen from
colleagues who satisfied the following criteria: (1) relatively low-pitched voice (to facilitate
interpretation of spectograms) and (2) fluent speech. All spoke variants of General
American English. Although all speakers had spent long periods living in different parts of
the US (mainly Midwest and Northeast), they had lived continuously in the South for the
last twenty years. All speakers used fully rhotic productions. Speakers were all between ages
53 to 63 at the time of the data collection. It would be difficult to ascribe a strong regional
accent to any speaker, although Speaker 4 had a more Southern accent than the others,
mainly evident in more centralized onset for the diphthong /o /. Specific profiles are as
follows. Speaker 1 (described in Oxley et al., 2006) grew up in west Pennsylvania in an
immigrant community where Russian was still spoken widely by many adults. His
fundamental frequency was characteristically very low, and at times dipped into the pulse
register. Speakers 2 and 4 were from the Midwest, and Speaker 3 from the East Coast
(Washington, DC area). Speaker 2 lived in the Michigan area. Speaker 4 noted that he had
grown up in several Midwest locations, his parents spoke with a rural Arkansas dialect.
Stimuli
Forty sentences were constructed such that tokens of /l/ were located in four (4) different
metrical/syllable (M/S) environments, and two (2) vocoid environments (front, /i/ and back,
/o / (see Table I for specific names and abbreviations used in the paper). The /o / vocoid was
chosen because variations in production (e.g. /o/) still yield a mid central back vocoid with
some degree of rounding. We acknowledge that the two intervocalic contexts receive different
degrees of stress, in that only the iambic context received tonic accent for the utterance, but
the foot structure is unaffected. Foot assignment followed the left-to-right scanning procedure
described by Liberman and Prince (1977). Certainly, higher levels of prosodic organization
(phonological or prosodic word, phonological phrase, and so forth) will potentially influence
the specific units and boundary strengths, particularly in the iambic contexts, but the
difference will be a matter of degree (Keating, Cho, Fougeron, & Hsu, 2003).
Contextual variability in American English dark-l 529
Table II. /l/ values averaged over 4 speakers, 4 prosodic/syllable contexts (# denotes word boundaries), and 2
vowel contexts (V denotes tautosyllabic vowel, and v, the reduced vowel)
Onsets Codas
Prosodic/Syllable Iambic Edge Iambic Trochaic Trochaic Edge
Context Miss#Cv Intervocalic vCV Intervocalic VCv VC#for
Vowel Front Back Tot Front Back Tot Front Back Tot Front Back Tot All
F1 /l/ Hz 367 388 378 381 382 381 453 399 426 438 417 427 404
F2 /l/ Hz 920 830 875 805 790 797 870 738 804 831 692 761 809
F2-F1 /l/ Hz 552 442 497 424 408 416 417 340 378 392 274 333 406
Duration /l/ sec .078 .075 .077 .110 .110 .110 .073 .076 .075 .085 .077 .081 .086
Intensity /l/ dB 69 69 69 68 67 68 70 70 70 69 68 69 69
530 J. Oxley et al.
therefore, measurements on only the first four productions of Speakers 2, 3, and 4 were
completed for the present study.
Statistical analysis. The first author entered the data into an SPSS 11.0 file. First, a
repeated measures MANOVA (four repetitions, four levels of context, and two levels of
vowel) was completed for each of the /l/ variables (F0-F3, and intensity). The multivariate
F-values for repetition effects alone and with the other variables were not significant, and so
data were collapsed across repetitions and the resulting means were used for subsequent
analyses. These averaged means reflect stable and typical values for the speaker (Table I).
The MANOVA was rerun, but without the repeating component. Post-hoc Tukey HSD
comparisons (p5.05) were completed on context following a significant effect. For the
context and vowel interaction, separate one-way ANOVAs on front vowel and back vowel
contexts, respectively were completed, together with Tukey HSD post-hoc comparisons to
analyse the relative performance of each context within each vowel environment. For
intervocalic contexts, univariate 4 (speaker)62 (IV contexts)62 (vowel) ANOVAs were
also completed for duration values of intervocalic vCV and VCv segments. A separate
ANOVA was completed for the rimes in the two /l/-coda contexts.
Results
Dialect: Dark-l Use
We collected speech samples of vowels in hVd contexts, and /l/ embedded in citation forms
(e.g. ili, ala, ulu, olo) and used the F2 values of /l/ to compare our speakers’ data with data
reported in Recasens (2004) and Lehiste (1964). Our speakers’ data were directly
comparable with that of the dark-l speakers recorded by Recasens, and with the American
speakers recorded by Lehiste (1964, cited in Recasens, 2004). Mean F2 values of /l/ in
citation form for all four of our speakers fell solidly within the range 1200–1300Hz.
Because our findings reflect a pattern of a dark /l/-onsets by all speakers, we used the second
set of hypotheses.
Formant values for the experimental stimuli appear in Table II, and confidence intervals
are in Figure 1. Note that although F2 values for the citation contexts were well within the
ranges reported separately by Lehiste and Recasens (i.e. between 1000–1300Hz), the
majority of values of /l/ in the sentence-embedded experimental stimuli were much closer to
800Hz. This pattern is entirely consistent with Huffman, who also embedded stimuli in
connected speech. She reported at least a few values within this range, and drew attention
to her finding that /l/ F2s were generally lower in connected speech than in citation form
stimuli. Degree of F2 lowering was very much a speaker effect in her data. Espy-Wilson
(1992) also documented some rather low F2 values, particularly in coda positions (between
998Hz), but her data also reflected much lighter tokens in onset positions. Her data,
however, were based on essentially citation forms.
Statistical findings
Analysis of variance was completed on F2, F1, F2-F1, and intensity values to evaluate the
influence of speaker, prosodic/syllable context (referred to hence as ‘‘context’’), and vowel
(back/front) on /l/. Formant and intensity values were taken at midpoint of the steady state
period. The results appear on Table III. We report partial eta squared (gp2) as a strong
index of power, and eta squared (g2) as a way to evaluate the relative contributions of each
effect (Tabachnick & Fidell, 1989). Figures 2–4 show g2 values for F1, F2, and F2-F1 /l/.
Contextual variability in American English dark-l 531
Speaker effect
Overall, the Speaker effect was the most powerful for all variables. Tukey HSD comparisons
(p,.05) indicated that mean F2 values for the speakers showed significant differences as
follows: Speaker 2 (715 Hz),Speakers 1 and 4 (801, 815Hz, respectively),3 (907Hz).
There was less than 200Hz separating the speakers’ mean values. Interaction effects
(Table III) generally reflect these absolute differences among the speakers. For F2-F1, each
speaker differed significantly from the others: 2 (322Hz),1 (349Hz),4 (422Hz),3
(532Hz). Differences in F1 thus differentiate Speakers 1 and 4. Increases in F1, which we
may assume result from dorsum retraction/lowering toward the pharyngeal wall, contribute
to darkening in [l]. From high to low, speakers’ values were 1 (451).2 (393) and 4 (393).3
(375). Speaker 1 (see Oxley et al., 2006) differed considerably from the others in terms of the
high F1 evident in his productions. Speaker variation (i.e. idiolect) is not the focus of this
paper, and we will not address speaker effects further in this paper. We note, however, that
although the speakers differed from one other in absolute terms, they generally presented the
same pattern of variation of dark-l according to the contextual factors of interest.
Context effect
Figure 5 presents mean [l] formant values for the four prosodic/syllable contexts pooled for
both vowels. When F2-F1 is used as a measure of /l/ colouring, we find that the four
contexts differ significantly from one another, with the following Hertz values (darker to
lighter): TEC (331),TIVC (378),IIVO (416),IEO (497). For the F2 measure, there
532 J. Oxley et al.
Partial eta
Effect df F Sig. squared 1 Eta squared 2
Notes: 1. From SPSS 11.0 printout (reflects independent contribution of each effect in relation to total error SS;
values do not add to 1). 2. Calculated by hand using SS Effect/ Corrected Total from SPSS 11.0 printout: allows
for comparison of relative contributions of effects (added values (1).
was also a strong main effect for context (Table II), and post-hoc Tukey HSD (p,.05)
comparisons indicated significant differences between mean F2 values for all pairs except
the two intervocalic contexts IIVO and TIVC. Changes in F1 account for the significant
difference between the two intervocalic contexts observed in the F2-F1 measure. F1 values
separated themselves into codas (430 and 426) and onsets (381 and 378). Thus, in
intervocalic contexts, F1 strongly determines any further darkening of dark-l, and F1
raising in the rime accounts for the difference. The contribution of F1 to l-darkening is seen
in elevated F1 values for the two coda contexts (TIVC, TEC). Tukey HSD post-hoc
comparisons (p,.05) were significant for the difference between l-onsets and l-codas. In
terms of our hypotheses, the trend of declining values was supported.
Vowel effect
The main effects on the three measures were all significant (Table II) and in the expected
direction. Front vowel contexts yielded lighter l-tokens in terms of (a) the F2-F1 measure
Contextual variability in American English dark-l 533
Figure 5. Mean F1, F2, F2-F1 values pooled for vowel context for the four prosodic/syllable contexts IEO, IIVO,
TIVC, and TEC.
(i.e. ‘‘That’s peel/pole for me’’), the effect of the closing oral cavity in anticipation of /f/
would further lower the F2 value.
The noteworthy finding is that the iambic intervocalic onset context produced a much
lower F2 value for /l/, relative to all other front vowel contexts. Part of the explanation is
that this context represents a between-foot boundary, and is thus stronger than its within-
foot, trochaic intervocalic counterpart. Stronger boundaries permit complete gestures. The
difficulty with boundary strength alone is that it cannot account for distribution of F2
values in the back vowel contexts. An alternate account, and one we prefer, is that front
vowel contexts generally prevent the darker-l tokens generally unless there is a strong
prosodic boundary and a preceding vowel with which dark-l may freely coarticulate (e.g. a
reduced vowel). We note that the iambic intervocalic context (That’s a leaf/loaf…) also
changes the /l/-boundary in that if falls within a phonological word (v) ‘‘a leaf’’, leaving the
/l/ as an unfooted syllable directly under v head. A comparison of Figures 6 and 7 reveals
that the strong boundary context allows for F2 lowering, but /l/ colouring is also determined
by F1 raising, which serves to darken /l/ codas in back vowel contexts.
Duration
We looked first at the relationship between the duration (sec.) of cVC and CVc intervals in
relation to /l/ formant values. A 4 (speaker)62 (intervocalic contexts)62 (vowels) ANOVA
536 J. Oxley et al.
was completed. Unsurprising significant effects for speaker and vowel effects were noted;
however we report only the variable of interest, namely main effect of context. The mean
duration in seconds for IIVO interval (.331) was significantly longer than the trochaic
counterpart (.280), F (1, 64)532.801, p,.001, and the increased IIVO duration within the
front vowel context provided the environment for a relatively dark /l/ (Figures 1, 6 and 7).
Boundary effects can account adequately for the observed duration difference (i.e. the IIVO
context represents a between-foot boundary, whereas the TIVC context is foot-internal).
The longer duration of the IIVO vocalic interval may also be an indicator of tonic stress.
Prosody
Increased F0 and intensity are potential markers of stress. Because the target intervals in
the trochaic intervocalic context were utterance initial and did not receive tonic stress, we
compared (1) the intensity of vowels in carrier words across both metrical and vowel
contexts, and (2) the relative intensity of carrier words with the non-carrier words that had
received tonic stress. For the first measure, a 4 (speaker)64 (context)62 (vowel) ANOVA
was completed on intensity values for the vowel tautosyllabic to /l/. Speaker and speaker
interaction effects were unsurprisingly significant, but there were no significant effects for
context, vowel, or the context by vowel interaction. We conclude that although context
IIVO received tonic stress, its intensity value was not different from context TIVC, which
received utterance-initial stress. Intensity values for vowels in key locations were made from
Contextual variability in American English dark-l 537
a random selection of utterances for which intonational contours were printed. Locations of
interest were: for IEO, the first two words ‘‘Say Miss’’; for IIVO, the first word ‘‘That’s’’,
for TIVC ‘‘peach’’ (etc.; see Table I for utterances). We then compared intensity values for
carrier words with words in comparable utterance positions. For all speakers, the first word
in utterances was loudest or at least of equal in intensity to the stressed word in IIVO. In the
phrase ‘‘Say Miss Leap’’ Miss always ranked third after the initial word and the carrier
word. For the back vowel counterpart, the carrier word and Miss were generally equally
loud. This reflects a pattern of greater intensity for the more open back vowels. For context
IIVO, the first word that’s was generally louder relative to the carrier word (That’s a leaf/
loaf), but sometimes it was of equal intensity. For context TIVC, the sentence initial carrier
word was always louder than the word that corresponded to carrier words in other
utterances (i.e. feel louder than peach). We conclude that, in terms of intensity level, the
difference between the position of the carrier-word in the TIVC and the other contexts was
not a factor. The privileged position of utterance-initial words (Feel a peach) was at least
equivalent, if not greater in intensity, to carrier words in the tonic stress bearing IIVO
stimuli (that’s a leaf).
F0 is also a measure of stress, and so F0 values were included in the multivariate
ANOVA reported earlier. We compared mean F0 values of the tautosyllabic vowel in the
carrier words across context and vowel, but found no significant effects for context, or
context by vowel. To examine the possibility that sentence position of carrier words might
influence production of /l/, we compared F0 contours. The observed pattern confirmed
538 J. Oxley et al.
Figure 8. Mean duration of intervocalic vCV and VCv intervals in front and back vowel contexts.
that for context TIVC, carrier words (feel a) exhibited falling contour that began at a lower
pitch than the tonic word (feel a peach for me). In context IIVO (that’s a leaf for me), the F0
dipped from schwa into l, and rose again in /i/ to a level equivalent to the word that’s.
Speaker preference for a rather trochaic cadence and relative emphasis on the first word
meant that although the carrier word in TIVC was not in the tonic position, it nevertheless
received comparable emphasis in terms of F0.
Discussion
The data strongly supported our predictions concerning the way dark-l would pattern in the
four speakers of American English. We note that all four speakers used characteristically
dark-l tokens, whether in citation form or in connected speech, and that connected speech
yielded F2 values that ranged from 200–400Hz lower than citation forms. Given this
pattern, it is of interest to investigate how contextual variations yielded the observed range
of F2 and F2-F1 values.
Our first prediction for dark-l onset speakers was that [l] would be darker in back vowel
than in front vowel contexts, and this unsurprising finding was supported. We predicted
that in back vowel contexts, /l/ would exhibit systematic F2 decline from canonical onset,
through intervocalic onset and coda positions, to canonical coda position, and this
progression was evident in all speakers, and supports the claim by Recasens (2004) that
differences in /l/ are scalar. Our third prediction, also supported across all four speakers,
Contextual variability in American English dark-l 539
Figure 9. Mean duration of edge CV and VC intervals in front and back vowel contexts.
was that in front vowel contexts, lowest F2 values for /l/ would occur in the iambic
intervocalic onset context. Last, we expected to find systematic interplay between F2 and
F1 in the form of F1 raising to effect darker codas when F2 was already low.
Increased F2 values (lighter tokens) were observed by characteristically dark-l onset
speakers in high front vowel edge contexts (i.e. when /i/ is tautosyllabic), and also in the
intervocalic within-foot boundary environment. The iambic intervocalic ‘‘schwa # [li]’’
context allowed for a between-foot boundary, a long vCV duration, and a shared dorsal
place feature +back between dark-l and the preceding schwa. The combined properties of
this particular context thus allow strong gestural affiliation between the dorsal feature of
dark-l and the preceding reduced vowel (schwa). A fully dark-l, comparable in terms of F2
and F2-F1 values from the corresponding back-vowel context, is facilitated.
Evidence of the syllable position of dark-l in back vowel contexts seems to be evident
mainly in F1 behaviour. We found greatly increased F1 values for dark-l in coda positions.
F1 raising reflects post-dorsum retraction (pharyngeal constriction) (Pickett, 1999). With
F2 already very low in back vowel contexts, additional darkening of codas is realized by F1
raising, which results in a lower F2-F1 measure. The vowels in both rimes (intervocalic and
edge) differed significantly from the vowels in l-onset syllables. It is well known that coda-l
‘‘shades’’ the preceding vowel. Preliminary findings from an ongoing perceptual
investigation of the influence of onset dark-l on the reduced vowel suggests that l-
colouring, in the form of the +back feature, is evident early in the vCV signal when listeners
540 J. Oxley et al.
are asked to identify sounds present in a series of sound clips developed by a forward
truncation process (Roussel & Oxley, 2006).
We know from the duration data that the IIVO vCV interval was much longer than its
TIVC VCv counterpart, but our current data set is not set up to evaluate how much of this
durational difference was due to tonic stress prominence and how much to the foot- or
other boundaries. Still, it seems clear that a relatively dark-l could not be produced in the
front vowel TIVC context. Duration of the sonorant interval containing dark-l with a
tautosyllabic high front vowel is an important determinant of F2.
In tautosyllabic back vowel/diphthong contexts, the dorsal gesture of dark-l
coarticulates freely with the tautosyllabic vowel/diphthong. There is not gestural conflict,
and so timing is no longer an issue. Syllable position and boundary strength determine l-
colouring.
We acknowledge that carrier words for the intervocalic intervals were not situated in the
same part of the utterance, but a review of F0, F0 contour, and intensity values of vowels
tautosyllabic to /l/ suggested that their location did not result in differences in either
intensity or F0 difference. Only duration could be a clue as to the status of stress in the
utterance. This finding is consistent with the effect of domain-initial strengthening reported
by Keating et al. (2003). As we reported in Oxley et al. (2006), the particular strengthening
in the case of dark-l occurs for the dorsal gesture.
Catford (2001) remarked that, to an RP speaker, GAE can have a pharyngeal quality that
comes in part from an overall colouring from the ‘‘bunched’’ /r/ characteristic of many GAE
speakers. The dorsal ‘‘bunching’’ and retraction gesture contrasts with the /r/ that is produced
with apical elevation. Productions of the word ‘‘Colorado’’ by many speakers of GAE can
illustrate the pharyngeal quality that Catford notes: it is loaded with segments that share the
feature +low. Gick et al. (2002) advocate the need to incorporate ‘‘a more efficient model of
tongue root specification into articulatory models of speech production, and for a unified
treatment of the tongue root component of liquids and vowels in theories of phonology’’
(p. 370). From our data, we present acoustic evidence (i.e. F1 raising) that supports a strong
pharyngeal component of dark-l. Connected speech introduces numerous other variables
which potentially modulate the overall colouring of /l/. These include prosody, rate of speech,
boundary strength, and coarticulatory effects of other consonants and vowels in terms of
place, manner, voicing. Carter and others have argued that l-ness is evident across much wider
domains within an utterance than adjacent vowels. It seems plausible that degree of
pharyngeal constriction, reflected in F1, may be an important factor that determines the
overall quality of an English dialect. Further research is needed to determine this conjecture.
For dialects with a distribution of dark-l onsets, our data suggest that clinicians may find
it fruitful to investigate whether their /l/-gliding clients do so in particular contexts; for
example, phrases of the form determiner + an l-initial word (a/the/my leaf), where the word
contains a tautosyllabic high front vowel. Determiners tend to be unstressed or reduced in
contextual speech, except for contrastive usage. The determiner + word context thus sets
up the phonetic context for an exaggerated dorsal gesture, and potential lip rounding,
which may actually increase the speaker’s tendency to glide. This is a primary context for
word-initial, but phonetically intervocalic /l/ in popular therapy activities, such as ‘‘Go
Fish!’’, in which the speaker must repeatedly request items containing a target sound
(‘‘Have you got a/the leaf?’’). Furthermore, from a perceptual perspective, the presence of a
strong +back dorsal feature, together with a slow acoustic transition, more characteristic of
a glide, may influence which features are most salient to the listener. Thus, the clinician’s
model of such a phrase may not be particularly contrastive to the child. Our data suggest
Contextual variability in American English dark-l 541
only that post-reduced-vowel words with a dark-l onset and a tautosyllabic high front vowel
might be problematic, but not what would be better. We speculate that use of words with
lower front vowels may help, because there would not be a conflicting context for the dorsal
gestures of /l/ and the vowel. The critical point is that stress patterns, syllable shape, and
surrounding vowels exert a synergistic influence on /l/ production and must be considered
carefully in the selection and development of clinical material. These wider contextual
concerns are finding their way into clinical- and clinical teaching literature (e.g. Bauman-
Waengler, 2004; Gordon-Brannan & Weiss, 2007; Kamhi & Pollock, 2005; Smit, 2004;
Velleman, 1998), but there continues to be a lag between the identification of this complex
array of variables in the literature, and its routine use in clinical practice, where there
continues to be dependence on a more narrow and simplistic three-place (initial, medial,
final) model of context.
Acknowledgment
We wish to thank an anonymous reviewer for helpful suggestions made regarding the
presentation of information. This research was supported in part by the Louisiana
Experimental Program to Stimulate Competitive Research (EPSCoR), funded by the
National Science Foundation and Board of Regents Support Fund.
References
Ball, M. J., & Müller, N. (2005). Phonetics for communication disorders. Mahwah, NJ: Lawrence Erlbaum.
Bauman-Waengler, J. (2004). Articulatory and phonological impairments, second edition. Boston, MA: Pearson
Education, Inc.
Bladon, R. A. W., & Al Bamerni, A. (1976). Coarticulation resistance in English /l/. Journal of Phonetics, 4,
137–150.
542 Contextual variability in American English dark-l
Bladon, R. A. W., & Nolan (1977). A video-fluorographic investigation of tip and blade alveolars in English.
Journal of Phonetics, 5, 185–193.
Carter, P. (2003). Extrinsic phonetic interpretation: spectral variation in English liquids. In J. Local, R. Ogden, &
R. Temple (Eds.), Phonetic Interpretation: Papers in Laboratory Phonology VI (pp. 237–252). Cambridge:
Cambridge University Press.
Catford, J. C. (2001). A practical introduction to phonetics, second edition. Oxford: Oxford University Press.
Espy-Wilson, C. (1992). Acoustic measurements for linguistic features distinguishing the semivowels /w j r l/ in
American English. Journal of the Acoustical Society of America, 92, 736–757.
Gick, B., Kang, A., & Whalen, D. H. (2002). MRI evidence for commonality in the post-oral articulations of
English vowels and liquids. Journal of Phonetics, 30, 357–371.
Giles, S. B., & Moll, K. L. (1975). Cinefluorographic study of selected allophones of English /l/, Phonetica, 31,
206–227.
Gordon-Brannan, M. E., & Weiss, C. E. (2007). Clinical management of articulatory and phonological disorders, third
edition. Baltimore, MD: Lippincott, Williams & Wilkins.
Grunwell, P. (1987). Clinical phonology, second edition. Baltimore, MD: Williams & Wilkins.
Hardcastle, W., & Barry, W. (1989). Articulatory and perceptual factors in /l/ vocalisations in English. Journal of
the International Phonetic Association, 15, 3–17.
Huffman, M. K. (1997). Phonetic variation in intervocalic onset /l/’s in English. Journal of Phonetics, 25, 115–141.
Johnson, W., & Britain, D. (2003). L Vocalisation as a naturally occurring phenomenon. Essex Research Reports in
Linguistics, 44, 1–37, Accessed 2 July 2006, available at: http://www.essex.ac.uk/linguistics/wkpaps/index.shtm.
Kamhi, A. G., & Pollock, K. E. (2005). Phonological disorders in children. Baltimore, MD: Paul H. Brookes.
Keating, P., Cho, T., Fougeron, C., & Hsu, C. (2003). Domain-initial articulatory strengthening in four
languages. In J. Local, R. Ogden, & R. Temple (Eds.), Phonetic Interpretation: Papers in Laboratory Phonology VI
(pp. 145–163). Cambridge: Cambridge University Press.
Lehiste, I. (1964). Acoustic characteristics of selected English consonants. (Bloomington IN: Indiana University
Research Center in Anthropology, Folklore and Linguistics.
Liberman, M., & Prince, A. (1977). On stress and linguistic rhythm. Linguistic Inquiry, 8, 249–336.
Narayanam, S., Alwan, A., & Haker, K. (1997). Toward articulatory models for liquid approximants on MRI and
EPG data. Part I. The laterals. Journal of the Acoustical Society of America, 101, 1064–1077.
Oxley, J. D., Buckingham, H., Roussel, N., & Daniloff, R. (2006). Metrical/syllabic factors in English allophony:
Dark /l/. Clinical Linguistics and Phonetics, 20, 109–117.
Pickett, J. M. (1999). The acoustics of speech communication. Boston, MA: Allyn & Bacon.
Recasens, D. (1995). Velarization degree and coarticulatory resistance for /l/ in Catalan and German. Journal of
Phonetics, 23, 37–52.
Recasens, D. (2004). Darkness in [l] as a scalar phonetic property: implications for phonology and articulatory
control. Clinical Linguistics and Phonetics, 18, 593–603.
Recasens, D., & Espinosa, A. (2005). Articulatory, positional and coarticulatory characteristics for clear /l/ and
dark /l/: evidence from two Catalan dialects. Journal of the International Phonetic Association, 35, 1–25.
Recasens, D., & Pallares, M. D. (2001). Coarticulation, assimilation, blending in Catalan consonant clusters.
Journal of Phonetics, 29, 273–301.
Roca, I., & Johnson, W. (1999). A course in phonology. Oxford: Blackwell.
Roussel, N., & Oxley, J. D. (2006). Context Effects on Production of Dark /l/: Perceptual Evidence. Poster
presented at the 2006 Convention of the American Speech Language Hearing Association, Miami Beach,
Florida.
Shine, R. E. (1989). Articulatory production training: A sensory-motor approach. In N. A. Creaghead, P. W.
Newman, & W. A. Secord (Eds.), Assessment and remediation of articulatory and phonological disorders, second
edition (pp. 335–359). New York: Macmillan.
Shriberg, L. D., & Kent, R. D. (2003). Clinical phonetics, third edition. Boston, MA: Allyn & Bacon.
Smit, A. B. (2004). Articulation and phonology resource guide for school-age children and adults. Clifton Park, NY:
Delmar Learning.
Sproat, R., & Fujimura, O. (1993). Allophonic variation in English /l/ and its implications for phonetic
implementation. Journal of Phonetics, 21, 291–311.
Tabachnick, B. G., & Fidell, C. S. (1989). Using multivariate statistics, second edition. New York: Harper Row.
Velleman, S. L. (1998). Making phonology functional. Boston, MA: Butterworth-Heinemann.
Webber, S. L. (2006). Artic Photos Fun DecksH For Articulation & Language. Greenville, SC: Super Duper
Publications.
Wells, J. (1982). Accents of English. New York: Cambridge University Press.