Further analysis of the articulation of /r/ in Polish - The postconsonantal position

Abstract

The present paper constitutes an extension of the research presented in Stolarski (2013), where the articulation of the Polish /r/ was tested in the intervocalic position. It investigates the postconsonantal articulation of the rhotic, the second context in which, according to Biedrzycki (1978: 83-84), the segment may be articulated as a tap instead of the trill. In order to accomplish this task, 10 respondents from different parts of Poland were instructed to read into a microphone 26 natural Polish words in which /r/ was located after consonants. All these articulations were analysed acoustically in terms of the number of constricted intervals in the pronunciation of the consonant under discussion and the data obtained were assessed statistically. In addition, some other measurements, such as the length and relative strength of the intervals, were taken. They provide further information on the nature of the articulation of /r/ in Polish.

Full text

SKY Journal of Linguistics 28 (2015), 349–379
Łukasz Stolarski
Further Analysis of the Articulation of /r/ in Polish –
the Postconsonantal Position
Abstract
The present paper constitutes an extension of the research presented in Stolarski (2013),
where the articulation of the Polish /r/ was tested in the intervocalic position. It
investigates the postconsonantal articulation of the rhotic, the second context in which,
according to Biedrzycki (1978: 8384), the segment may be articulated as a tap instead
of the trill. In order to accomplish this task, 10 respondents from different parts of
Poland were instructed to read into a microphone 26 natural Polish words in which /r/
was located after consonants. All these articulations were analysed acoustically in terms
of the number of constricted intervals in the pronunciation of the consonant under
discussion and the data obtained were assessed statistically. In addition, some other
measurements, such as the length and relative strength of the intervals, were taken. They
provide further information on the nature of the articulation of /r/ in Polish.
1. Introduction
This section briefly summarises the universal characteristics of rhotics and
common assumptions regarding the articulation of the Polish /r/. Moreover,
the results of the articulation of the rhotic in the intervocalic position
presented in Stolarski (2013) are summarised and the aims of the present
study are clearly defined.
1.1 Rhotics as a diverse group
  -        
constitute a highly diverse group of sounds. The International Phonetic

UKASZ STOLARSKI
350
Alphabet1 lists at least 8 symbols which belong to this class, but the
possible phonetic realisations of rhotics lie on a phonetic continuum
involving many manners and places of articulation. Ladefoged and
Maddieson (1996: 217) mention that rhotics are realised as trills, taps,
    -   
articulation may combine features of several of these groups. Nevertheless,
Maddieson (1984) observes that in the descriptions of rhotics in various
          
   r     
approximants or fricatives constitute only 13.5% of the cases.
As far as the place of articulation is concerned, Maddieson (1984)
claims that in the descriptions of rhotics the dental/alveolar region is
mentioned most frequently; however, it is rarely defined precisely. He
additionally suggests that other phonetic realisations involve retroflexes
(12%), dentals (3.2%), uvulars (0.9%) and palatao-alveolars (0.6%).
These general statistics should be treated with caution. Reports on the
phonetic realisation of rhotics in various languages differ in their level of
detail and, in some cases, the assumption that a segment is realised in a
given way does not fully correspond to the actual articulation. A good
example of this is the extensive discussion on the production of the Spanish
phonemic trill. Standard descriptions indicate that the segment is realised
with two to five constricted intervals (Recasens & Pallarès 1999).
Nonetheless, the rhotic has been reported not to be articulated in the
          number of
dialects (Rissel 1989; Diez-Canseco 1997; Hammond 1999; Adams 2002;
Hualde 2005; Bradley, Colantoni & Willis 2006; Díaz-Campos 2008). In
fact, production of the phoneme has been used in the categorisation of
Spanish dialects (Resnick 1975; Lipski 1994).
A possible common acoustic feature present in all rhotic articulations
was proposed by Ladefoged (1975) and Lindau (1978). They suggested that
in approximant realisations, as well as in vowel-like segments separating
strictures in trills, the third formant is lowered. A similar tendency can also
be observed in transitions to and from rhotic consonants. Nevertheless,
Ladefoged and Maddieson also provide many counterexamples and
    
1 <http://www.internationalphoneticalphabet.org/ipa-charts/ipa-symbols-chart-
complete/>

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
351
historical connections between these subgroups, and on the choice of the
Ladefoged & Maddieson 1996: 245).
Since trilling and tapping are assumed to be the two major phonetic
realisations of rhotics, it is worth mentioning that there is an important
inconsistency in the way the two manners of articulation are defined. On
the one hand, a clear distinction is made by the number of the constricted
intervals: one in the case of taps and more than one in trills (Lindau 1985;
Gimson & Cruttenden 2001; Dhananjaya, Yegnanarayana & Bhaskararao
2012). On the other hand, authors such as Ladefoged and Maddieson
(1996) and Laver (1994) propose an alternative interpretation, according to
which trilling does not necessarily have to involve actual constrictions. In
the production of these sounds it is the position of the articulators in a
configuration which potentially allows vibrations which is crucial, rather
than the actual presence of the vibrations. Ladefoged and Maddieson
(1996: 217
  
realisation of a sound. In the current paper, Ladefoged and 
interpretation is accepted. Nevertheless, an additional discussion on the
problems resulting from such disparate definitions will be presented in
Section 4.
1.2 Assumptions about the articulation of the Polish /r/
In the literature on Polish phonetics the consonant /r/ is consistently
referred to as an alveolar trill. As summarised in Stolarski (2013), its other
allophones include the palatalised [rj] (Benni 1959: 28; Wierzchowska
1967: 157, 1971: 138, 1980: 119; 
Dukiewicz 1995: 40;   ; Gu   
 
] (Benni 1959: 29; Wierzchowska 1971: 122, 1967: 105;
     ; Dukiewicz 1995: 40), the
uvular // (Wierzchowska 1971: 122; Pilich 1975: 51;  
132; Dukiewicz 1995: 40; Gussman 2007: 27),

j] (Benni 1959: 29), and also the uvular fricative [], or even the dental
approximant [] (Gussman 2007: 27). Additionally, some authors mention a
possible tapped realisation of the consonant (Wierzchowka 1967: 122,
1980: 73; Jassem 1973: 246
Szczepankowski 1985: 47; Dukiewicz 1995: 40), but such an articulation is
supposed to be less typical than trilling.

UKASZ STOLARSKI
352
Among the factors which may lead to a tapped pronunciation of the
Polish /r/, some authors propose the     
Jassem 1981). It is presumed that in slow speech the consonant tends to be
articulated as a trill; however, the faster the speech becomes, the more
likely it is for the phoneme to be realised as a tap. Moreover,

is frequently encountered when /r/ follows /i/ or /j/ and the consonant is
palatalised (although full palatalisation of the rhotic is, generally, a
questionable claim and slight palatalisation of the Polish /r/ is more
realistic). Dukiewicz (1995: 40) and Jassem (1974: 60), on the other hand,
contend that in casual, non-emphatic speech the tapped [] may be
encountered in any phonetic context. Finally, Biedrzycki (1978: 8384)
suggests that even though the Polish /r/ is basically a trill, especially before
consonants and in the coda position, as in kartka dar
  ped articulation may be encountered between vowels and
after consonants, e.g. in zaraz     strój 
(Biedrzycki 1978: 83).2 The experiment described in Stolarski (2013) was
  ulation of the
consonant in the intervocalic position as one of the contexts in which
tapping could be expected. This paper is an extension of the previous
analysis and focuses on the second context suggested by Biedrzycki  the
postconsonantal position. Here, again, some articulations of /r/ may involve
tapping rather than trilling, but no research has ever been done to verify
this assumption.
It is worth mentioning that from the acoustic point of view, /r/,
regardless of whether it is realised as a tap or a trill, involves vocoidal
elements which are interrupted by one, in the case of tapping, or more, in
the case of trilling, constricted intervals. Jassem (1973) calls these vocoidal
 -    and Maddieson (1996)
refer t-
  
and they appear before and after each constricted interval (Pilich 1975;
Wierzchowska 1980).  (1983: 118122) underlines the fact that the
-
even though the speaker is usually unaware of their presence during
2 As far as the posconsonantal position of /r/ is concerned, Biedrzycki (1978) does not
specify whether or not the rhotic must be tautosyllabic with the preceding consonant for
tapping to occur, but the results presented in Section 3.3 do not indicate that such a
condition is necessary.

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
353
articulation. One of the reasons why they appear in the production of trills
and taps could be that voiceless constrictions may be more effectively
distinguished in contrast to these vocoidal elements.
  -  
the constricted intervals, but their precise duration may depend on
particular characteristics of rhotics in a given language. For example, they
are estimated to last approximately 50 milliseconds in Italian (Ladefoged &
Maddieson 1996: 219), but only 2030 milliseconds in Polish (Jassem
1973: 245246). Furthermore, they have clearly a discernible formant
structure which closely resembles the acoustic configuration found in the
central vowel // (Wierzchowska 1967: 105; Jassem 1973: 245246;
Szczepankowski 1985: 47).
It must be stressed that the acoustic characteristics measured in the
experiment described in Section 3 concern the nature of constricted
intervals. They are central to the current paper because their behaviour is
decisive in the classification of rhotics. Nevertheless, a detailed description
  -         
important task which could be undertaken in future research.
1.3 Results of the preceding study
The major aim of the experiment reported in Stolarski (2013) was to
investigate the most typical phonetic realisation of the Polish /r/ in the
intervocalic position in terms of the number of constricted intervals. In
order to achieve this goal an acoustic analysis of the articulation of the
rhotic was conducted. The test materials involved 18 natural Polish words
in which the consonant was pronounced between vowels. The words
represented all the possible intervocalic contexts in Polish. The articulation
of the consonant was measured between high, low, front, central and back
vowels.
There were five men and five women taking part in the experiment.
All of them were native speakers of Polish and they came from different
parts of Poland. The participants were asked to read into a microphone the
18 words in the most natural way. In order to avoid any possible bias
resulting from the conscious articulation of the consonant, the participants
were not informed as to what was being investigated and the examples
were mixed with random Polish words.

UKASZ STOLARSKI
354
All the recordings were analysed in Praat (Boersma & Weenink 2011)
and each articulation of /r/ was evaluated in terms of the number of
constricted intervals. The final results of the experiment are quoted below:
Table 1. Final results of the experiment reported in Stolarski (2013)
Trilled stop
Tapped stop
Weak tap
with friction
Sum in all VrV contexts
5
171
4
Percentage
2.8%
95%
2.2%
Margin of error
2.4%
3.2%
2.2%
Surprisingly, trilled articulation was encountered in less than 3% of the
cases. The tapped allophone [] was expected to be found in the
intervocalic context, but the frequency of its occurrence was astounding,
nevertheless. Indeed, the results of the experiment proved that in the
investigated phonetic context the Polish /r/ is realised almost exclusively as
a tap. This questioned the predominant opinion expressed in the literature
on Polish phonetics that the consonant is basically a trill. Obviously, the
conclusion reached in Stolarski (2013) concerns only the intervocalic
position and more research needs to be done in order to fully explore the
actual nature of the consonant under discussion. The present publication
examines the articulation of the Polish /r/ in the postconsonantal position,
which is the next necessary step.
Other conclusions reached in Stolarski (2013) include:
1. No relationship between any of the phonetic characteristics of the vowels
surrounding the tested consonant and the manner in which it is articulated was
observed. Tapping was equally frequent between high, low, front, central and
back vowels.
2. The average length of the (first) constricted interval seemed to depend on both
the vertical and horizontal articulatory dimensions of the surrounding vowels. It
was the longest between front vowels (
 = 25.9 ms, sd = 7.9 ms, n = 40), shorter
between central vowels (
= 21.8 ms, sd = 5.5 ms, n = 40) and the shortest
between back vowels (
= 19.2 ms, sd = 3.7 ms, n = 40). The results for the
contrast high versus low were 
= 23.9 ms, sd = 7.8 ms, n = 50 and 
= 20.4 ms,
sd = 4.6 ms, n = 70, respectively.
3. The vocoidal element preceding the (first) constricted interval was, on average,
louder than the one following it. (The vocoidal elements were measured only in
terms of voice intensity in order to establish the strength of the (first) stricture;
they were distinguished from the preceding and following vowels by transitions in
formant structure.)

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
355
4. The strength of the closure depended on the horizontal articulatory dimension
of the surrounding vowels  it was the weakest between back vowels and the
strongest between front vowels.
5. There was a tendency for the tapped articulation to be slightly less frequent
among female participants than among male participants. Still, the difference was
not large and the sample used in the experiment was too small to prove this
observation.
6. Men tended to pronounce /r/ with, relatively speaking, weaker contact between
the tongue blade and the alveolar ridge than women.3 Again, the sample was too
small to statistically prove this observation.
1.4 Aims of the present study
This study aims at examining the most typical realisation of the Polish /r/ in
the postconsonantal position. The prime objective is to establish the typical
number of periods of reduced acoustic energy present in the articulation of
the consonant. In other words, this paper aims at answering the question
whether the Polish /r/ is realised as a trill or a tap when it is pronounced
after other consonants. The choice of this particular phonetic environment
    (1978) suggestions summarised in Section 1.2.
Since his claim that the Polish /r/ happens to be realised as a tap in the
intervocalic position has been validated in Stolarski (2013), examining the
postconsonantal articulation of the consonant is a natural follow-up.
Analogously to the sub-aims pursued in Stolarski (2013), in the
present experiment the approximate average length of the constricted
intervals will also be estimated. Even though the results obtained in
Stolarski (2013) supported the assumption made   (1983: 119)
-
is still important to measure the length in another phonetic environment.
Additionally, the relative strength with which the contacts between the
tongue blade and the alveolar ridge is made will be measured, but the
method applied is slightly different from the one used in Stolarski (2013).
The new phonetic context examined in this paper necessitates changes in
estimating the statistic (see the antepenultimate paragraph in Section 2.3).
3 The relative strength with which the contact was made was established on the basis
of the differences between the maximum intensity in the vocoidal phases surrounding
the (first) constriction and the minimum intensity during the (first) constriction.

UKASZ STOLARSKI
356
2. Methods
This section describes the participants and the materials used. Furthermore,
the procedures employed in designing and performing the experiment are
presented and a detailed discussion on the types of articulation found in the
data, as well as the criteria for their differentiation, is provided.
2.1 Participants
The informants taking part in the experiment are the same native speakers
of Polish who were selected in Stolarski (2013). Out of the total ten
respondents, five are male and five are female and their ages vary from 20
to 33 years old. Three men and three women come from the area around the
city of Kielce in central-eastern Poland. Additionally, one female
respondent comes fro
the eastern part of Poland around Lublin. The only difference between the
current group of participants and the one recorded in the previous
experiment is that instead of the second person who lives in the Lublin
region, a 23-year-old man from Warsaw was asked to take part in the test.
It is worth adding that in Stolarski (2013) no correlation between any
measured statistics and the place of origin of individual participants was
found. Nevertheless, it is still important to analyse recordings of
respondents coming from different regions of Poland in case any
differences in their articulation of /r/ happen to emerge in the current
experiment.
2.2 Selection of materials
The selection of materials was based on the principle that the articulation of
/r/ should be tested after various consonants. In fact, the most viable
solution seemed to be to find examples with /r/ following every single
Polish consonant. This would enable the testing of the possible articulatory
influence of the place and manner of articulation of neighbouring segments.
It is plausible that there may be variations in the phonetic realisation of /r/
within the postconsonantal context itself.
There are many proposals on the exact number of consonantal
phonemes in Polish, but entering this considerable debate is beyond the
scope of the present article. The set which has finally been chosen is based

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
357

all other accounts and their phonemic status is rarely questioned. The
inventory is summarised below:
 stops: bilabial /p, b/, post-dental /t, d/, velar /k, g/, velo-
 affricates: alveolar /
, 
/, post-alveolar /
,
/, alveolo-palatal / 
,
/
 fricatives: labio-dental /f, v/, post-dental /s, z/, post-alveolar /, /, palatal /,
velar /x/
 trills: alveolar /r/
 approximants: labio-velar /w/, palatal /j/
 nasals: bilabial /m/, post-dental /n/, palatal //
 laterals: alveolar /l/
Among the consonants which have been excluded from the analysis are /c,
/. They have highly limited distribution and appear only before /i, j/ and
//, so the segment under discussion never follows any of them. In addition,
the articulation of /r/ will not be tested after another /r/, since trilling in
such a context is the only predicted phonetic realisation.
It is worth mentioning that in most studies on the consonant system of

phonetic point of view, testing the pronunciation of /r/ after such a
phoneme would be equivalent to testing it in the word-initial environment,
and this is not the aim of the present publication.
All in all, the test materials involved 27 examples (see below). Most
of them are common in everyday Polish. In two cases (Czrezpienianie,
wadźrajana) rare proper names were chosen because the intended sound
combinations /
/ and /
r/ could not be found in any popular words.
Moreover, the examples differ in terms of the position of the rhotic within
the syllable. According to the criteria for syllable division proposed by
Gimson and Cruttenden (2001: 5152) seventeen involve /r/ which is
tautosyllabic with the preceding consonant (examples 17, 13 and 1725)
and in the remaining ten cases the rhotic is in a different syllable than the
preceding segment (examples 812, 1416, 2627). One of the possible
influences of these differences will be discussed in Section 3.3.
It should be noted that the vowels which follow the rhotic in the
examples are random. No particular pattern was chosen because the results
of the previous experiment described in Stolarski (2013) indicated that the
phonetic characteristics of vowels surrounding /r/ do not affect its
articulation type; they only influence additional attributes, such as the
relative strength and length of the (first) closure (see Section 1.3), which
are of secondary importance to the present study.

UKASZ STOLARSKI
358
The words used in the experiment:
(1) prawy 
(2) brat 
(3) trawa 
(4) drogo 
(5) Kraków 
(6) grono 
(7) Czrezpienianie (the name of a medieval Slavic tribe)
(8) brydż rodzinny 
(9) lubić rowery 
(10) wadźrajana (a Buddhist movement)
(11) socrealizm 
(12) telewidz rozbawiony 
(13) mrówka 
(14) Konrad 
(15) koń rasowy 
(16) bal roku 
(17) fragment 
(18) wrona 
(19) srebro 
(20) zranić 
(21) szrama 
(22) żrący 
(23) środa 
(24) źródło 
(25) chrapka 
(26) półroczny -
(27) chojrak 
As in the experiment described in Stolarski (2013), in order to avoid
ous articulation of /r/, the
examples have been mixed with other words. To be more precise, half of
the items on the list presented to the participants were the examples quoted
above and the other half were random words used in modern Polish.
Consequently, it was very difficult for the informants to guess the aim of
the test.
2.3 Procedure
The participants were presented with a list of words. It is important to
stress that the words (or, in some cases, pairs of words) were written in
isolation. If they were to be produced within a sentence frame, the phonetic

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
359
processes occurring in connected speech could reduce the chance of the
trilled articulation appearing. Producing the forms in isolation was most
         educed forms
typical of connected speech.
The instructions were to read the list into a microphone in the most
natural way.
As in the experiment reported in Stolarski (2013), the recordings were
analysed in Praat (Boersma & Weenink 2011) and the following
measurements of /r/ were taken:
 the type of articulation
 the duration of the (first) closure
 the drop in intensity during the (first) closure
The first property was measured in order to establish the type of
articulation of the Polish /r/ in the postconsonantal position (see Section
1.4). The two additional acoustic characteristics were included in this study
because they provide supplementary data on the way in which trilling and
tapping is realised and allow additional conclusions to be drawn. Moreover,
they were also discussed in the previous study on the intervocalic
articulation of the Polish /r/ and a comparison between current and the
former results referring to the two attributes could help to provide
additional explanation regarding the nature of the constrictions in the
rhotic.
In the present analysis four types of articulation could be observed.
The first three were also encountered in the intervocalic position discussed
in Stolarski (2013). Definitions and examples of the four types are provided
below:
1. A trilled stop  in this case the production of /r/ involves more than
one closure, visible as lighter areas in spectrograms and periods of smaller
acoustic activity in oscillograms. It is interesting to note that in the present
experiment more than two constricted intervals were only observed in one
articulation out of 270. Namely, the word mrówka 
with three taps by Informant 3. In all other cases of trilling there were only
two closures detected. An example of such an articulation is given in
Figure 1.

UKASZ STOLARSKI
360
Figure 1. Articulation of wrona 
(the two closures are marked by arrows)
2. A tapped stop  in this case the production of /r/ involves one definite
closure. An example is provided in Figure 2.
Figure 2. Articulation of chrapka 
tapped /r/ (the closure is marked by an arrow)
3. One (weak) closure with an immediate intensification in higher
frequencies  in Stolarski (2013) such an articulation was treated as a
special sub-type of the tapped stop. According to the definition adopted
there, simultaneously with the tap is an intensification of acoustic energy in
the higher regions of the spectrum. It is useful to treat this type of
articulation separate because it is rather different from a prototypical tap
and its constricted interval is less distinct (see Figure 3).4 Consequently,
4 The interval indicated by the arrow in Figure 3 could also be interpreted as a very
short approximant because the formant structure of the surrounding vocoidal elements is
partially retained. However, the short length of this approximation and the

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
361
measuring the duration of the stricture in such cases involved not only the
visual inspection of spectrograms and oscillograms, but also judgements
based on auditory information.
Figure 3. Articulation of źródło Type 3.
losure is indicated by an arrow
4. A trilled fricative / a fricative / an approximant  this type involves
various articulations in which no definite closure could be detected. In
some cases minor fluctuations in intensity combined with intensification of
acoustic energy in higher regions of the spectrum were observed, which
suggests slight, rapid movements of the tongue tip combined with friction.
Consequently, such phonetic realisation of /r/ may be referred to as a trilled
fricative, or a fricative trill (Laver 1994: 264; Trask 1996: 151). In other
cases articulation lacking any discernible closure involved only friction. An
example of this is presented in Figure 4. Additionally, the constricted
interval was occasionally substituted by approximation which was open
enough for the friction not to occur. What makes all these articulations
distinct from the other categories described above is complete lack of the
constricted interval. Consequently, the segmentation of the elements in the
articulations which were classified as Type 5 included detailed auditory
analysis.
intensification of acoustic energy in higher frequencies make such articulations distinct
from prototypical approximants.

UKASZ STOLARSKI
362
Figure 4. Articulation of socrealizm 

The classification of the recorded articulations of /r/ was primarily based on
spectrographic analysis. Each pronunciation was compared with the four
prototypical types described above and categorised accordingly. It must be
stressed, however, that this classification has been established solely for
statistical evaluation. In order to prepare a statistical summary of the
problem, the continuum of possible phonetic realisations had to be
converted into countable, categorical data. As can easily be predicted, some
of the analysed recordings were not easy to classify, as they did not fall
neatly into any of the four categories. Problems with the distinction
between Type 3 and Type 4 were especially frequent, as the two kinds of
articulation are substantially similar. The only essential difference between
them is the presence of relatively weak constriction in the former, and lack
of any detectable closure in the latter. The decision to classify a given case
involved a very detailed analysis of visual data as well as auditory
judgements.
It should be noted that in the case of trilled articulation the
measurements of the approximate duration of closures concern only the
first period of reduced acoustic energy. The second (and third in the
pronunciation of mrówka b   -  
not analysed, because it was always less distinct and measuring its length
would yield inaccurate results. The duration of the first constriction,
nonetheless, was relatively easy to establish. As mentioned previously,
such constricted intervals are visible as lighter areas in spectrograms and an
additional inspection of oscillograms allows relatively precise
segmentation.

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
363
As far as the relative strength of the closure(s) in individual
articulations of /r/ are concerned, they may be estimated on the bases of
   -     
exhaled volume of air. The firmer the contact between the blade of the
tongue and the alveolar ridge, the greater the drop in intensity. This was

sequences [aa] and [aa]. In the former case, where /r/ was realised as a
tap, the drop in intensity was significant, but not in the latter, which
involved an approximant articulation. The method which has been adopted
in the present paper is similar to the one used previously, but certain
changes were necessary. In Stolarski (2013) the intensity shifts were
calculated as the difference between the minimum intensity value measured
in the articulation of the closures and the maximum value found in the
neighbouring vocoidal elements. In the present study, however, the
-
are preceded by vocoidal components which belong solely to /r/. These
-owels. In the
examples used in the present experiment /r/ follows only consonants.
Therefore, the drop in intensity will be established on the basis of the
difference between the maximum intensity value in the vocoidal
component preceding the closure, and the minimum intensity value
measured in the closure (see Figure 5). The vocoidal articulation following
-
in this case there is no clear-cut boundary between the vocoidal element
typical for the articulation of /r/ and the vowel.
Figure 5. Articulation of trawa        
depicted as a thin line in the middle of the spectrogram. The maximum intensity in the
vocoidal component preceding the closure and the minimum intensity in the closure are
marked by arrows

UKASZ STOLARSKI
364
It should be pointed out that the intensity shifts which have been measured
in this experiment seem to be strongly influenced by the voicing of the
preceding consonants (for a detailed explanation, see the penultimate
paragraph in Section 3.2). As a result, the shifts are not a good point of
reference for comparing the strength of constricted intervals in different
groups of examples arranged according to the place and manner of
articulation of segments preceding /r/. While in some groups both voiced
and voiceless consonants are included, in other cases only voiced segments
precede the rhotic. For this reason, direct comparisons are not possible.
Nevertheless, the intensity shifts measured in the experiment are useful in
          
articulation of the consonant (see Section 3.4).
As far as the statistical methods used in Section 3 are concerned, z-
tests were applied for the comparisons involving categorical data. In the
case of numerical data, t-tests were performed where the sample size was
smaller than 30, and z-tests were applied for samples larger than 30.
3. Data analysis
The analysis of the results includes general observations which refer
directly to the major aims of the paper (Section 3.1) as well as a detailed
discussion on the effects of the place of articulation, manner of articulation,
and voicing of the consonants preceding /r/ in the articulation of the rhotic
(Section 3.2). Furthermore, in Section 3.3 the problem of a possible
influence of the position of /r/ within the syllable is analysed and Section
3.4 examines additional, extralinguistic factors which may have influenced
the results.
3.1 Results
The overall results of the present test are reported in Table 2. They show
that trilling, referred to as Type 1 (see Section 2.3), is extremely rare in the
pronunciation of /r/ in the postconsonantal position in Polish. It was
encountered in only 4 out of 270 articulations and the 95% confidence
interval for its occurrence in the population of speakers of standard Polish
is 1.48% ± 1.44%. It is interesting to note that the four observed cases of
trilled /r/ ware pronounced by only one participant. Her articulation of the
consonant in the words wrona  szrama   chrapka
        

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
365
mrówka 
result is very similar to the one obtained in the former experiment designed
to study the intervocalic context (see Table 1), where the consonant under
discussion was trilled in 2.8% of the cases. As a consequence, the claim
made in most studies on Polish phonetics that the typical articulation of the
Polish /r/ involves trilling needs to be reconsidered. It has been thoroughly
refuted in the present paper as well as in Stolarski (2013). Studying the
preconsonantal and word-final articulation of /r/ is the natural next step. As
summarised in Section 1.2, Biedrzycki (1978: 8384) predicted that the
consonant may happen to be realised as a tap in both intervocalic and
postconsonantal positions, but the fact that trilling was found to be so
infrequent is astonishing. His other assumptions suggest that in the
preconsonantal and word-final phonetic environments /r/ should be realised
as a trill, but in the light of the evidence gathered so far other phonetic
realisations seem to be possible even in these contexts.
The present experiment reveals that the most typical articulation of /r/
in Polish in the postconsonantal environment is tapping (Type 2). It was
observed in almost 80% of the cases, and the 95% confidence interval for
such an articulation among the population of speakers of standard Polish is
79.63% ± 4.8%. A tapped stop may be regarded as the basic allophone of
/r/ in the postconsonantal position as well as in the formerly studied
intervocalic environment. The corresponding result obtained in Stolarski
(2013) revealed that [] was produced in 95% of the recorded articulations.
Type 3, which involves a single weak closure with an immediate
intensification of acoustic energy in higher frequencies, was found in
12.22% of the cases, which is more frequent than in the intervocalic
environment investigated in the former experiment. As it was mentioned in
Section 2.3, however, this manner of articulation could also be regarded as
a special kind of tapping. If its occurrence was calculated together with
Type 3, tapped articulation would constitute over 90% of all the cases
analysed in the present experiment.
Phonetic realisations which do not involve any definite closure,
referred to as Type 4, were observed 18 times, which constitutes 6.67% of
the cases. They are less common than the ones grouped under Type 3 (and,
obviously, Type 2), but still more frequent than instances of trilling, which
are exceedingly rare.

UKASZ STOLARSKI
366
Table 2. General test results
Type 1
Type 2
Type 3
Type 4
Sum in all Cr contexts
4
215
33
18
Percentage
1.48%
79.63%
12.22%
6.67%
Margin of error
1.44%
4.8%
3.91%
3.03%
Duration of
the (first)
closure
Maximum
intensity in
the vocalic
element
preceding the
(first) closure
Minimum
intensity
during the
(first) closure
Intensity drop
during the
(first) closure
Average in all Cr
contexts
20.7 ms
76.2428 dB
72.6999 dB
3.5428 dB
Standard deviation
7.4 ms
4.1159 dB
4.1201 dB
2.7893 dB
Margin of error
0.878 ms
0.490957 dB
0.491458 dB
0.332716 dB
The average duration of the closures (or various kinds of approximations in
Type 4) amounts to 20.7 milliseconds. It is slightly shorter than the mean
length 21.9 milliseconds observed in Stolarski (2013), but on the bases of
the samples analysed in the two experiments this difference is statistically
non-significant (p = 0.0651). It is interesting to note that although the
distribution of individual measurements here is approximately normal, it
involves some outliers which drive the average upward. This is readily
visible in Graph 1 which presents the histogram of individual closure
durations with 30 intervals spaced 2 milliseconds apart ranging from 3.2
milliseconds to 62.5 milliseconds. Because the distribution is slightly
skewed to the right, the median, which is 19.3 milliseconds, probably better
reflects the average duration of the closures.
The average intensity drop in the production of /r/ in the intervocalic
position reported in Stolarski (2013) was greater than the one observed in
the current experiment. The difference which amounts to 2.3963 ± 0.4414
dB is highly statistically significant (p < 0.0001); however, it should be
borne in mind that the methods of measuring the intensity shifts in the two
experiments differed (see Section 2.3). Therefore, the discrepancy between
the two results does not necessarily prove that the contact between the
blade of the tongue and the alveolar ridge in the production of /r/ tends to
be firmer between vowels than after consonants, although such a tendency
is possible.

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
367
Graph 1. Histogram of individual closure durations with 30 intervals spaced every 2
milliseconds
3.2 Influence of the place of articulation, manner of articulation and
the voicing of the preceding consonants
The summary of the articulation of /r/ after different groups of consonants
divided according to the manner of articulation (see Table 3) suggests that
trilling is encountered almost exclusively after fricatives, but it must be
stressed that the relative frequency of Type 1 is still very low (3.33%).
More importantly, the trilled /r/ occurred in the pronunciation of only one
participant (see Section 3.1), so the conclusion that trilling is more frequent
after fricatives cannot be statistically substantiated by the results of the
present experiment. The fact that Types 3 and 4 are more common after
fricatives and affricates is, however, evident. The articulatory
characteristics of these two groups of consonants cause the following /r/ to
frequently be produced with either a weak closure accompanied by
immediate intensification in higher frequencies or friction substituting any
definite constricted interval. In other phonetic contexts /r/ was produced
almost exclusively as a prototypical tap.

UKASZ STOLARSKI
368
Table 3. Articulation of /r/ after different classes of consonants grouped according to
the manner of articulation
Type 1
Type 2
Type 3
Type 4
Duration of
the (first)
closure or
approximation
Intensity
drop
during the
(first)
closure
After stops (on the basis of 60 articulations)
Sum
0
60
0
0
Average
20 ms
3.6916 dB
Percentage
0%
100%
0%
0%
Standard
deviation
5.5 ms
2.3854 dB
Affricates (on the basis of 60 articulations)
Sum
0
34
15
11
Average
23.3 ms
3.7132 dB
Percentage
0%
56.67%
25%
18.33%
Standard
deviation
10.5 ms
3.1940 dB
After nasals (on the basis of 30 articulations)
Sum
1
27
2
0
Average
20.5 ms
4.1571 dB
Percentage
3.33%
90%
6.67%
0%
Standard
deviation
5.4 ms
2.5546 dB
After laterals (on the basis of 10 articulations)
Sum
0
10
0
0
Average
20.5 ms
3.8419 dB
Percentage
0%
100%
0%
0%
Standard
deviation
4.7 ms
2.8634 dB
After fricatives (on the basis of 90 articulations)
Sum
3
66
15
6
Average
19.5 ms
2.9704 dB
Percentage
3.33%
73.33%
16.67%
6.67%
Standard
deviation
5.9 ms
2.7779 dB
After approximants (on the basis of 20 articulations)
Sum
0
18
1
1
Average
22.8 ms
4.8396 dB
Percentage
0%
90%
5%
5%
Standard
deviation
8.4 ms
2.5050 dB
The average duration of the (first) closure was slightly longer after
affricates and approximants than in other phonetic contexts and these
differences are in most cases statistically significant. For instance, the two-
tailed p-value for the difference between 20 milliseconds after stops and
23.3 milliseconds after affricates is 0.031. Nonetheless, it must be pointed
out that the articulation of /r/ after affricates was tested on either fairly rare
Polish words or on the border between two words. These factors might
-like

syllable structure, see Section 3.3).
The differences between the values of intensity drop during the (first)
closure summarised in Table 3 are not statistically significant in most

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
369
cases. The p-values are below 0.05 only when we compare the result
obtained after fricatives (2.9704 dB) to the ones after nasals (4.1571 dB)
and approximants (4.8396 dB). In these cases the two tailed p-values are
0.0312 and 0.0031, respectively.
Table 4. Articulation of /r/ after different classes of consonants grouped according to
the place of articulation
Type 1
Type 2
Type 3
Type 4
Duration of
the (first)
closure
Intensity
drop during
the (first)
closure
After bilabial (on the basis of 40 articulations)
Sum
1
39
0
0
Average
18.2 ms
3.1222 dB
Percentage
3%
98%
0%
0%
Standard
deviation
5.1 ms
2.0719 dB
After labio-dental (on the basis of 20 articulations)
Sum
1
17
1
1
Average
16.5 ms
1.1549 dB
Percentage
5%
85%
5%
5%
Standard
deviation
3.9 ms
2.2007 dB
After post-dental (on the basis of 60 articulations)
Sum
0
58
2
0
Average
20.4 ms
3.9907 dB
Percentage
0%
97%
3%
0%
Standard
deviation
4.7 ms
2.3081 dB
After alveolar (on the basis of 20 articulations)
Sum
0
11
7
2
Average
21.2 ms
2.7411 dB
Percentage
0%
55%
35%
10%
Standard
deviation
11.4 ms
2.7964 dB
After post-alveolar (on the basis of 40 articulations)
Sum
1
18
9
12
Average
23.6 ms
3.7899 dB
Percentage
3%
45%
23%
30%
Standard
deviation
9.1 ms
3.1132 dB
After alveolo-palatal (on the basis of 20 articulations)
Sum
0
16
3
1
Average
23.1 ms
4.6900 dB
Percentage
0%
80%
15%
5%
Standard
deviation
9.4 ms
3.3798 dB
After palatal (on the basis of 40 articulations)
Sum
0
27
11
2
Average
22.5 ms
4.4900 dB
Percentage
0%
68%
28%
5%
Standard
deviation
7.3 ms
3.0441 dB
After velar (on the basis of 30 articulations)
Sum
1
29
0
0
Average
20.7 ms
3.4770 dB
Percentage
3%
97%
0%
0%
Standard
deviation
6.1 ms
2.5967 dB

UKASZ STOLARSKI
370
Table 4 presents the results of the experiment arranged according to the
place of articulation of the consonants preceding /r/. It is readily apparent
that Types 3 and 4 tend to be more frequent after segments which are
alveolar, post-alveolar, alveolo-palatal and palatal. Type 3 and Type 4 were
found in 25% and 14.17% of these cases, respectively. Examples in which
the rhotic is preceded by consonants with other places of articulation
comprise 2% for Type 3 and 0.67% for Type 4. These large differences are
highly statistically significant (in both cases the p-values are smaller than
0.0001).
There are two possible explanations for such a tendency. Firstly, the
Polish /r/ is an alveolar consonant. If it is preceded by a segment which has
the same or similar place of articulation, its production will naturally
become more strenuous than in cases when it is preceded by consonants
articulated at some other location in the mouth. For example, it is obvious
that the sequence /
r/ is harder to pronounce than /kr/ or /pr/. In the former
case the speaker needs to place the tongue on the alveolar ridge and make a
complete closure, behind which the air pressure builds up. Then she needs
to release the air with an accompanying friction typical for affricates and
immediately after that, with the same part of the tongue, perform the action
of tapping (or trilling). In the latter cases the blade of the tongue is not
involved in the articulation of the preceding consonants and the production
of /r/ is easier. The speaker has more control over the movements of the
blade of the tongue and the closure stage is performed with more precision.
This interpretation is substantiated by the differences in the mean duration
of the closures. The constricted intervals were comparatively longer when
/r/ was pronounced after alveolar, post-alveolar, alveolo-palatal and palatal
consonants than in other cases. It is probably also the result of the
articulatory positions allowing less control over the movements of the
blade of the tongue. After consonants with a place of articulation around
the alveolar region, the constricted intervals are pronounced with less
precision.
The second explanation is that it may be not the place but rather the
manner of articulation which triggers the phonetic realisations of Type 3
and 4. In fact, all Polish alveolar, post-alveolar and alveolo-palatal
consonants are either affricates of fricatives (see Section 2.2). In addition,
two out of the four palatal phonemes are fricatives (/ and //). Such a
convergence of phonetic characteristics makes it difficult to decide on the
conditioning of the tendency. One possible conclusion would be just to
state that /r/ is more frequently articulated with friction or approximation

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
371
after affricates and fricatives which are pronounced around the area of the
alveolar ridge.
Table 4 suggests that intensity drops during the (first) closure may be
correlated with the duration of the stricture. The longer the period of
reduced acoustic energy, the greater the drop in intensity. Nonetheless, this
tendency is much less evident in Table 3. Consequently, in order to
establish the strength of the correlation all measurements of the durations
of individual closures have been compared with the corresponding drops in
intensity and the correlation coefficient obtained is 0.53. This indicates that
the positive linear relationship is not very strong. A scatterplot of all the
durations of individual closures against the corresponding shifts in intensity
is shown in Graph 2.
Graph 2. Scatterplot of the durations of individual closures against the corresponding
shifts in intensity (r = 0.53)
Table 5. Articulation of /r/ after different classes of consonants grouped according to
voicing
Type 1
Type 2
Type 3
Type 4
After voiceless (on the basis of 110 articulations)
Sum
2
85
10
13
Percentage
1.82%
77.27%
9.09%
11.82%
After voiced (on the basis of 160 articulations)
Sum
2
130
23
5
Percentage
1.25%
81.25%
14.38%
3.13%

UKASZ STOLARSKI
372
Duration of
the (first)
closure
Maximum
intensity in
the vocalic
element
preceding the
(first) closure
Minimum
intensity
during the
(first) closure
Intensity drop
during the
(first) closure
After voiceless (on the basis of 110 articulations)
Average
20.4 ms
74.8133 dB
72.1864 dB
2.6270 dB
Standard
deviation
7.9 ms
4.1204 dB
4.0628 dB
2.7317 dB
After voiced (on the basis of 160 articulations)
Average
21.4 ms
77.0927 dB
72.7794 dB
4.3133 dB
Standard
deviation
6.9 ms
3.8375 dB
4.1768 dB
2.5973 dB
Another factor which influences the values of intensity shifts is the voicing
of the consonants preceding /r/ (see Table 5). In this context the
relationship is rather close. The difference between the mean values of
intensity drops after voiceless consonants and after voiced consonants is
1.6863 ± 0.65 dB (p < 0.0001). It must be stressed, though, that the mean
minimum intensity values during the closure are very similar in the two
contexts. The difference results mainly from the dissimilar values of the
average maximum intensity in the vocoidal element preceding the
constricted interval. They differ by 2.2794 ± 0.9729 dB and the observation
is, again, highly statistically significant (p < 0.0001). The natural
conclusion is that the type of voicing of the preceding consonants has a
direct effect on the level of intensity of the vocoidal element articulated
before the closure. When /r/ is pronounced after voiced segments, the
vocoidal element tends to be louder than when /r/ is articulated after
voiceless segments. All this points to the fact mentioned in Section 2.3 that
intensity shifts measured in this experiment should not be treated as a
reliable point of reference for comparing the strength of constricted
intervals in the different groups summarised in Tables 3 and 4. The values
which have been obtained are biased by the unequal duration of the
closures and by the type of voicing of the preceding consonants.
The data reported in Table 5 do not reveal any important differences
in the frequency of occurrence of the types of articulation of /r/ defined in
Section 2.3. Trilling is equally rare after voiceless and voiced segments and
other kinds of phonetic production do not seem to vary substantially. Only

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
373
Type 4 should be treated as slightly more frequent after voiceless
consonants than after voiced consonants (p = 0.0049).
3.3 Tautosyllabic versus heterosyllabic /Cr/ sequences
The /Cr/ sequences in seventeen examples analysed in the present
experiment are tautosyllabic (examples 1, 2, 3, 4, 5, 6, 7, 13, 17, 18, 19, 20,
21, 22, 23, 24 and 25). In the other ten cases the rhotic and the directly
preceding consonant are heterosyllabic. The two groups are compared in
Table 6.
The differences in the type of realisation of the rhotic in the two
groups are very small. Type 2 and Type 4 are distributed in an almost
identical manner. The largest difference concerns Type 3, but even in this
case it is too small to be statistically relevant (4.4% ± 8.39%, p = 0.2846).
These results suggest that the type of articulation of the rhotic does not
depend on whether or not the consonant preceding /r/ and /r/ are
   (1978) assumption that the
tapped realisation of the rhotic may be encoutered in the postconsonantal
position (see Section 1.2) may be extended to heterosyllabic sequences.
As far as the length of the stricture is concerned, it tends to be longer
in heterosyllabic sequences. The average difference between the two
groups amounts to 2.6 ms ± 1.74 ms and should be treated as statistically
significant (p = 0.0035). Additionally, in heterosyllabic clusters the mean
intensity drop during the (first) closure is larger than in tautosyllabic ones.
Again, this difference is statistically significant (0.9276 dB ± 0.6352 dB, p
= 0.0042).
Table 6. Articulation of /r/ according to whether or not the preceding consonant belongs
to the same syllable as the rhotic
Type 1
Type 2
Type 3
Tautosyllabic (on the basis of 170 articulations)
Sum
4
136
18
Percentage
2.35%
80%
10.59%
Heterosyllabic (on the basis of 100 articulations)
Sum
0
79
15
Percentage
0.00%
79.00%9
15.00%

UKASZ STOLARSKI
374
Duration of
the (first)
closure
Maximum
intensity in
the vocalic
element
preceding the
(first) closure
Minimum
intensity
during the
(first) closure
Tautosyllabic (on the basis of 170 articulations)
Average
19.9 ms
76.2459 dB
72.9911 dB
Standard
deviation
5.8 ms
3.6096 dB
3.8946 dB
Heterosyllabic (on the basis of 100 articulations)
Average
22.5 ms
76.2394 dB
72.0570 dB
Standard
deviation
7.7 ms
3.9694 dB
4.1624 dB
3.4 Influence of the informants’ characteristics: gender and place of
origin
The statistical summary of the results arranged according to the gender of
the respondents (see Table 7) suggests that men tend to articulate the Polish
/r/ slightly less clearly. In their pronunciation, Types 3 and 4 occur more
frequently, although on the basis of the sample used in the current
experiment these differences cannot be proven statistically. It is also
noticeable that the average duration of the (first) closure is somewhat
longer in male pronunciation. Even though the difference is statistically
significant (p = 0.0174), the corresponding result in Stolarski (2013) casts
doubt on its relevance. Namely, in the preceding experiment the average
duration of the closu
In the case of the results concerning shifts in intensity the tendency is
the same as in Stolarski (2013). The drop is greater in female speech.
Again, the sample is too small to prove that the difference is statistically
significant (p = 0.0912). Nonetheless, when we calculate the two-tailed p-
value on the basis of the data from both experiments the difference
becomes relevant (p = 0.0297). Therefore, it may be concluded that women
tend to articulate /r/ with stronger contact between the tongue blade and the
alveolar ridge than men.
The results of the present study do not reveal any crucial differences
which would depend on the place of origin of the participants. The same
observation was reported in Stolarski (2013).

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
375
Table 7. Results of the test arranged according to the gender of the participants
Type 1
Type 2
Type 3
Type 4
Results for female participants (on the basis of 135 articulations)
Sum
4
112
13
6
Percentage
2.96%
82.96%
9.63%
4.44%
Results for male participants (on the basis of 135 articulations)
Sum
0
103
20
12
Percentage
0.00%
76.30%
14.81%
8.89%
Duration of
the (first)
closure
Maximum
intensity in
the vocalic
element
preceding the
(first) closure
Minimum
intensity
during the
(first) closure
Intensity drop
during the
(first) closure
Results for female participants (on the basis of 135 articulations)
Average
19.8 ms
75.3541 dB
71.4706 dB
3.8835 dB
Standard
deviation
7.4 ms
3.7660 dB
3.6178 dB
2.8374 dB
Results for male participants (on the basis of 135 articulations)
Average
21.9 ms
77.1328 dB
73.8196 dB
3.3132 dB
Standard
deviation
7.1 ms
4.2930 dB
4.3015 dB
2.7092 dB
4. Discussion and conclusions
The overall results of the experiment (see Section 3.1) indicate that the
trilling of /r/ is extremely uncommon in Polish. Amongst the 270 analysed
articulations trilling was observed no more than 4 times and only in the
pronunciation of one informant. The most typical phonetic realisation of
the consonant is a tapped stop, which was articulated in 79.6% of the cases.
Such a result is similar to the one reported in Stolarski (2013), where the
production of /r/ was examined in the intervocalic context. All this data
leads to the conclusion that the widely accepted view that the consonant is
a trill needs a fundamental revision.
It is useful to focus at this point on the two definitions of trilling
 
         culation of the
Polish rhotic is most probably incorrect. Its basic allophone in the two
analysed phonetic environments is a tap and trilled articulation in these

UKASZ STOLARSKI
376
contexts is extremely rare. However, the final answer is less
straightforward if we consider     
        
  
(Ladefoged & Maddieson 1996: 217218). In such a case the claim that the
Polish rhotic is actually a trill is harder to dismiss. Nevertheless, the data
obtained in the experiment suggest that even according to the gestural
account the realisation of this consonant does not normally involve trilling.
Namely, the average duration of the closures reported in Table 2 (20.7 ms)
is shorter than the duration of trills in other languages. Taking into account
the statistics mentioned in Dhananjaya et al. (2012), Ladefoged and
Maddieson (1996) or Lever (1994), for the lingual gesture to be trill-like in
a one contact rhotic, the duration of the closed phase should be about 50 ms
or more.
A possible reason for the rhotic rarely being trilled might be a narrow
range of allowable variation of the aerodynamic conditions suitable for
trills (Ladefoged & Maddieson 1996; Solé, Ohala & Ying 1998; Solé
2002). If such small variations occur, the resulting articulation involves
various kinds of fricatives, approximants and, possibly, taps. In the
discussion on the production of the Spanish rhotics, briefly summarised in
Section 1.1, this factor is frequently quoted as an explanation for a number
of common deviations from trilling in numerous dialects.
Another aspect which is worth mentioning in the current discussion is
a frequent lack of phonemic contrasts between rhotics in a given language.
Maddieson (1984) suggests that the most typical distribution of liquids in
languages of the world involves one lateral and one r-sound. In the case of
languages with a larger number of liquids than two, there are usually more
laterals than r-sounds. This relatively small diversity of rhotics in a single
language has also been observed by Ladefoged and Maddieson (1996: 239)
who discuss Australian languages which usually have only two rhotics,
even though the contrasts among stops, nasals, and laterals involves four
          

237). All this suggests that any claims that a rhotic in a given language is
realised as a trill should be interpreted with caution, especially if a
language involves only one phonemic r-sound. If there is no contrast
           
makes trilling very unlikely, as seems to be the case in Polish. Saleh et al.
(2015) reach a similar conclusion in their analysis of the Egyptian Cairene

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
377
/r/: although it may be realised either as a trill or a tap, the latter articulation
is more frequent. Finally, it must be stressed that even a phonemic contrast
between trilling and tapping in a language does not guarantee that trilling
will actually be phonetically realised, as has been found in many dialects of
Spanish (see the discussion in Section 1.1).
Examining the other contexts mentioned by Biedrzycki (1978: 8384)
 the preconsonantal and word-final positions  is the next step. Even
though Biedrzycki states that in these two environments the Polish /r/
should be trilled, in the light of the current findings there is reason to
suspect that tapping will be observed in these contexts as well.
References
Adams, Catalina (2002) Strong Assibilation and Prestige: A sociolinguistic Study in the
Central Valley of Costa Rica. Unpublished doctoral dissertation. University of
California.
Benni, Tytus (1959) Fonetyka Opisowa Języka Polskiego [The Descriptive Phonetics of
Polish]. 
Biedrzycki, Leszek (1978) Fonologia Angielskich i Polskich Rezonantów [The
Phonology of English and Polish Sonorants].  we
Wydawnictwo Naukowe.
Boersma, Paul & Weenink, David (2011) Praat: doing phonetics by computer
[Computer program]. Version 5.2.03, retrieved 22 April 2011 from
<http://www.praat.org/>.
Bradley, Travis (2006) Phonetic realizations of /sr/ clusters in Latin American Spanish.
In Manuel Diaz-Campos (ed.), Selected Proceedings of the 2nd Conference on
Laboratory Approaches to Spanish Phonetics and Phonology, pp. 113.
Somerville, MA: Cascadilla Press.
Colantoni, Laura (2006) Increasing periodicity to reduce similarity: an acoustic account
of deassibilation in rhotics. In Manuel Diaz-Campos (ed.), Selected Proceedings
of the 2nd Conference on Laboratory Approaches to Spanish Phonetics and
Phonology, pp. 2234. Somerville, MA: Cascadilla Press.
Dhananjaya, N.; Yegnanarayana, Bayya & Bhaskararao, Peri (2012) Acoustic analysis
of trill sounds. Journal of the Acoustical Society of America 131 (4): 31413152.
Díaz-Campos, Manuel (2008) Variable production of the trill in spontaneous speech:
sociolinguistic implications. Selected proceedings of the 3rd Conference on
Laboratory Approaches to Spanish Phonology, pp. 4758. Somerville, MA:
Cascadilla Press.
Diez-Canseco, Susana (1997) Language Variation: The Influence of Speakers’ Attitudes
and Gender on Sociolinguistic Variables in the Spanish of Cusco, Peru.
Unpublished doctoral dissertation. University of Pittsburgh.
   Fonetyka polska [Polish Phonetics]. Warszawa/Kraków:


UKASZ STOLARSKI
378
Dukiewicz, Leokadia (1995) Fonetyka [Phonetics]. In Henryk Wróbel (ed.), Gramatyka
Współczesnego Języka Polskiego, Fonetyka i Fonologia [A Grammar of
Contemporary Polish: Phonetics and Phonology], pp. 9103. Kraków:

Gimson, Alfred Charles & Cruttenden, Alan (2001) Gimson’s Pronunciation of English.
Oxford: Oxford University Press.
Gussman, Edmund (2007) The Phonology of Polish. Oxford: Oxford University Press.
Hammond, Roberts (1999) On the non-occurrence of the phoneme /r/ in the Spanish
sound system. In Javier Gutierrez-Rexach & Fernando Martinez-Gil (eds.),
Advances in Hispanic Linguistics. Papers from the 2nd Hispanic Linguistics
Symposium, pp. 135151. Somerville: Cascadilla Press.
Hualde, José Ignacio (2005) The Sounds of Spanish. Cambridge: Cambridge University
Press.
Jassem, Wiktor (1973) Podstawy Fonetyki Akustycznej [Bases of Acoustic Phonetics].
Warszawa: 
 (1974) Mowa a Nauka o Łączności [Speech and the Communication Theory].

 (1981) Podręcznik Wymowy Angielskiej [Handbook of English Pronunciation].
wo Naukowe.
Ladefoged, Peter (1975) A Course in Phonetics. New York: Harcourt Brace Jovanovich.
Ladefoged, Peter & Maddieson, Ian (1996) The Sounds of the World’s Languages.
Oxford: Blackwell.
Laver, John (1994) Principles of Phonetics. Cambridge: Cambridge University Press.
Lindau, Mona (1978) Vowel features. Language 54 (3): 541563.
 (1985) The story of /r/. In Victoria Fromkin (ed.), Phonetic Linguistics: Essays in
Honor of Peter Ladefoged. pp. 157168. Orlando, FA: Academic Press.
Lipski, John (1994) Latin American Spanish. New York: Longman.
Maddieson, Ian (1984) Patterns of sounds. Cambridge: Cambridge university press.
Pilich, Jan (1975) Fonetyka Opisowa Języka Polskiego [The Descriptive Phonetics of
Polish]. agogicznej w Siedlcach.
Recasens, Daniel & Pallarès, Maria Dolors (1999) A study of // and /r/ in the light of
 coarticulation model. Journal of Phonetics 27: 14369.
Resnick, Melvyn (1975) Phonological variants and dialect identification in Latin
American Spanish. The Hague: Mouton.
Rissel, Dorothy (1989) Sex, attitudes, and the assibilation of /r/ among young people in
San Luis de Potosi, Mexico. Language Variation and Change 1: 26983.
   Zarys Fonologii, Fonetyki, Fonotaktyki i Fonostatystyki
Współczesnego Języka Polskiego [An Outline of the Phonology, Phonetics,
Phonotactics and Phonostatistics of Contemporary Polish]. 

Saleh, Marwa; Hegazi, Mona; Makram, Rania; Shoeib, Rasha; Sayed, Amal & Taher,
Sally (2015) Spectrographic analysis of Egyptian Cairene /r/: Is it a trill or a tap?
Logopedics Phoniatrics Vocology 40 (1): 3035.
Solé, Maria-Josep (2002) Aerodynamic characteristics of trills and phonological
patterning. Journal of Phonetics 30 (4): 655688.

FURTHER ANALYSIS OF THE ARTICULATION OF /r/ IN POLISH
379
Solé, Maria-Josep; Ohala, John & Ying, Goangshiuan (1998) Aerodynamic
characteristics of trills. Proceedings of the 135th Meeting of the ICA/ASA, pp.
29232924.
            
      Language, Thought and
Education: Across Systems, pp. 2135. Zielona Góra: Oficyna Wydawnicza
Uniwersytetu Zielonogórskiego.
 Zarys Fonetyki i Fonologii Języka Polskiego [An Outline of
the Phonetics and Phonology of Polish]. Kraków: Skrypty Uczelniane

Szczepankowski, Bogdan (1985) Fonetyka Akustyczna, Audytywna i Wizualna
Wizualna [Acoustic, Audiovisual and Visual Phonetics]. Warszawa:
Wydawnictwo Uniwersytetu Warszawskiego.
Trask, Robert Lawrence (1996) A Dictionary of Phonetics and Phonology. London:
Routledge.
   Opis Fonetyczny Języka Polskiego [The Phonetics of
Polish]. Warszawa: 
 (1971) Wymowa Polska [The Pronunciation of Polish].  

 (1980) Fonetyka i Fonologia Języka Polskiego [The Phonetics and Phonology of
Polish]. 
Willis, Erik (2006) Trill variation in Dominican Spanish: an acoustic examination and
comparative analysis. In Nuria Segarra & Almeida Jacqueline Toribio (eds.),
Selected Proceedings of the 9th Hispanic Linguistic Symposium, pp. 121131.
Somerville, MA: Cascadilla Proceedings Project.
Zarys Fonetyki i Fonologii Współczesnego Języka Polskiego
Polskiego [An Outline of the Phonetics and Phonology of Contemporary Polish].

Contact information:

Jan Kochanowski University

25-369 Kielce
Poland
e-mail: Lukasz(dot)Stolarski(at)ujk(dot)edu(dot)pl