Articulatory weakening as basis of historical rhotacism

Joaquín Romero, Sidney Martín Universitat Rovira i Virgili E-mail: [email protected]

includes under the same label a related phenomenon, ABSTRACT otherwise known as , by which dental/alveolar stops /t/ and/d/ surface as a tap or flap []. This process is Rhotacism is a well-known phonological process common in many dialects of English: American English: whereby dental/alveolar fricatives /s/-// develop into a pretty /prti/, get up /t p/ surface as [pri], [ p]; , most commonly a flap or tap. Results : thirteen /tin/ is often pronounced of a previous experiment suggested that the simple in manipulation of the temporal dimension can cause a [ ]. categorical shift in the perception from a fricative to a It seems that in both types of situations, classical flap. In this paper an experiment tested this further with rhotacism and flapping, the mechanisms at work should be speakers of American English, checking whether a the same, that is, a process of articulatory reduction that reduction in duration could change the percept from /z/ renders the same result: a fast, short tongue-tip movement to a flap. Results corroborate the previous findings. against the alveolar ridge that is accompanied by voicing. Some implications for the notions of gestural integrity are explored. The hypothesis that is presented in this paper is that the specific result obtained is due mostly to the reduction along the temporal dimension. It is suggested that a 1. INTRODUCTION sufficient amount of reduction in the duration of /s/-/z/ or /t/-/d/ will produce a flap-like segment that will sometimes Rhotacism is a common historical which has merge, though often just subphonemically, with already often been described as a change from a dental/alveolar existing flaps. fricative /s/—or more frequently the voiced counterpart /z/—to a rhotic alveolar sound that can be identified as a In a previous experiment [2] this hypothesis was tap or flap. Examples of historical rhotacism abound in investigated with Spanish and Catalan. Perceptual tests the literature: Latin genus, os (nominative) vs. generis, revealed that in both languages a simple reduction of oris (genitive); ceosan, leosan () vs. upwards of 50% in the duration of the fricatives /s/-/z/ was curon, luron (past). In these cases /s/-/z/ in intervocalic sufficient to give a percept of a flap. A clear case of position is replaced historically by an -like sound that can categorical perception was obtained, with speakers clearly be identified as an alveolar tap or flap. The result in these switching from the fricative to the flap once the duration cases is a phonological alternation that has morphological was reduced sufficiently. repercussions in the languages. These findings have implications for the notion of gestural Synchronic cases of rhotacism are also common, however. integrity, a relevant aspect within the theory of In some dialects of Castilian Spanish—most noticeably in Articulatory Phonology [3, 4, 5]. Articulatory gestures, by the Madrid area and in Northern parts of Castile—/s/ (or virtue of being the minimal units of phonological/phonetic its allophonic variation [z]) in coda position and organization, are integral units whose various components, preceding certain consonants is often replaced with an constriction location, constriction degree and stiffness alveolar flap. The resulting sound is identical to the (duration) cannot be manipulated separately. Thus, the alveolar flap which is the universal production of /r/ in reduction in the duration of a dental/alveolar fricative or Spanish, i.e. []. For example: doscientos [dosjentos], stop will have effects in the other two components of the dos veces [doz ees] can be pronounced [dojentos], gesture. It is precisely this situation that is hypothesized here for rhotacism: the reduction in the duration affects the [do ees]. Mallorcan Catalan also exhibits a strong constriction degree, from fricative/stop to flap, and the tendency to rhotacism in similar syllabic situations: location, at least in Spanish/Catalan, from dental to zmn z dit esmena [ ], es dit [ ] are often pronounced alveolar. As a consequence of the reduction, other [mn], [ dit]. associated gestures can also be altered, as in this case with the laryngeal gestures, which causes the change from Rhotacism has been described as a process of weakening voiceless /s/-/t/ to voiced []. [1], the result of articulatory relaxation and reduced effort. Exactly in what way a rhotic consonant like [] is a weak In this paper we report on the results of another version of /s/ or /z/ is not made clear, though. Hock also experiment that investigated the role of duration in the flapping process of American English, using the results of puzzle, puddle or something that was neither (other). Each synchronic data to explain aspects of historical sound subject listened to the complete set in one seating. change. We hypothesize that a reduction in the duration of /z/ in words like puzzle will result in the percept of a puzzle puzzle - 25% reduced /t/-/d/, in this case, puddle.

2. METHOD puzzle - 50% puzzle - 75% A perceptual experiment was devised that tested the hypothesis by presenting listeners with words embedded in carrier sentences. The design consisted of a set of target words that belonged to the following pairs: puzzle - 100% Table I. Experimental Design.

[VzV] [VV] Figure 1. Waveforms illustrating the manipulation of the lizard [lzd] littered [ld] VCV sequence from the original signal for puzzle. The razzle [ræzl] rattle [ræl] percentages indicate the part of the /z/ that was removed. The vertical lines isolate the remaining portion of the luz lu loser [ ] looter [ ] original /z/. causer [kz] caught her [k] 3. RESULTS puzzle [pzl] puddle [pl] The results were tabled and scored, and percentages of The target words were all bisyllabic with stress on the first identification were obtained for each set of tokens. Figures syllable. The target consonants were preceded by front 2 through 4 below illustrate the results obtained. In all vowels in the first two cases: high front in lizard/littered, these graphs the horizontal axis includes the 6 different and low front in razzle/rattle; in the next two cases there types of tokens that subjects heard, for examples, for were back vowels: high back in loser/looter and low back puzzle, they were: puzzle (the full, non-edited version of in causer/caught her; the last case included a central this word), puzzle-25 (the result of removing 25% of the vowel: puzzle/puddle. original word), puzzle-50 (the result of removing 50% of the original word), puzzle-75 (the result of removing 75% The original recordings were obtained from one male of the original word), puzzle-100 (the result of removing native speaker of General American English, who read 100% of the original word), and puddle (the full, non- each one of the words in Table I embedded in the carrier edited version of this word). This order is expected to sentence Tell me ______later twice. From each pair of represent a gradation in duration and in perception, from a tokens one was selected and used for acoustic full /z/ to a flap. manipulation before being presented to the listeners. Each of the target words with the [VzV] pattern—lizard, razzle, The vertical axis represents the percentage of times that subjects identified each one of those utterances as puddle, loser, causer, puzzle—was altered in the following way: that is, as containing a flap. Thus, a score of 0% indicates the acoustic portion corresponding to the middle that no speakers heard that token as being puddle, while a consonant was isolated and its duration was measured. score of 100% indicates that all speakers identified that Subsequently, the duration of the consonant was shortened token as being puddle, that is, as containing a flap. by 25%, 50%, 75% and 100% starting from the end of the segment. Effectively, a reduction of 100% was equivalent 100 100 to the complete removal of /z/. Figure 1 gives examples of 90 90 the resulting CVC sequences after the reduction. 80 80 70 70 60 60 % % 50 50 Each utterance was repeated once and the entire set, a total 40 40 of 60 tokens, was randomized for presentation to the 30 30 20 20 subjects. 10 native speakers of American English 10 10 participated in the experiment, which presented the 0 0 z material in an XYZ design, where X and Y showed a z -25 -50 -75 -100 -25 -50 -75 -100 [lzd] [ræzl] choice between a token with /z/ and a token with [], and Z corresponded to the other category. Thus, subjects heard Figure 2. Percentages of identification as the a token of, say, puzzle, with or without a manipulated corresponding token containing the flap for the pairs with signal, and they had to decide whether they had heard front vowels: lizard/littered and razzle/rattle. As the graphs in Figure 2 clearly illustrate, there is a sharp even after the removal of just 25% of /z/ at least one of the shift in the perception of both lizard and razzle tokens speakers identifies this as being puddle. The percentage of after the removal of 75% of the duration of /z/. In both identification increases steadily after removal of 50% and cases, 95% of the subjects heard these tokens as being 75%, even though at this level only 55% of the subjects littered and rattle, respectively. After the removal of 100% perceived puzzle as puddle. In this case, the categorical of the /z/, the percept of lizard was not distinguishable shift is again not so evident as for the front vowels, and from littered for all subjects. In the case of lizard even the rather similar to the effect obtained with the back vowels, removal of just 50% of the /z/ gives a 20% of especially causer/caught her. identification as littered. These results show a clear case of categorical shift in the perception of /z/ as a fricative or as 4. RESULTS a flap.

100 100 The results obtained in this experiment show a clear effect 90 90 of the shortening of /z/ in its perception as a fricative and a 80 80 70 70 robust trend to increase the perception as a flapped /t/ or 60 60 /d/ given a sufficient temporal reduction. In all cases and % 50 % 50 40 40 regardless of the vowel context, a temporal reduction of 30 30 20 20 100% in the duration of /z/ was sufficient to trigger the 10 10 0 0 almost universal perception of [ ]. With a level of

z -25 -50 -75 -100 z -25 -50 -75 -100 temporal reduction of 75% of the signal for /z/ in all cases [luz] [kz] the percept of [] reached over 50% and in two out of the five cases it reached 95%. Lower levels of Figure 3. Percentages of identification as the reduction—50%, 25%— were insufficient for most corresponding token containing the flap for the pairs with speakers to shift the category from /z/ to [], although in back vowels: loser/looter and causer/caught her. four out of the five cases at least some of the speakers felt The graphs in Figure 3 show a similar trend for the words that 50% was sufficient to cause the shift. with back vowels, although here the categorical shift is not as clear as with the front vowels. For loser, the clear break Though only from a qualitative point of view, there seems in the perception of /z/ between the removal of 50% and to be some effect of the vowel context on the percentage 75% is still present, but here even after removal of 100% of identification of /z/ as [ ]. For the front vowel context, of /z/ the percentage of identification as looter is only that is the pairs lizard/littered and razzle/rattle, the shift in 85%. In the case of causer, the trend is similar but here perception between /z/ and [] seems more obviously even more so than for loser, it seems that removal of 75% categorical than for either the back vowels or the central of /z/ is sufficient to give the percept of caught her for vowel. For these, there seems to be more of a gradient only 60% of the listeners. Still, after removal of 100% of progression of identification as [] as the reduction of the /z/, there is no hesitation in any speakers and they all signal gets more extreme. identify these as being tokens of caught her.

100 5. DISCUSSION 90 80 70 The data obtained from the present experiment are 60 % 50 interpreted as evidence that it is possible to explain 40 rhotacism as the result of a reduction in the articulatory 30 20 duration of a fricative like /z/. It has been proven that a 10 tongue-tip fricative like /z/ can be perceived as an alveolar 0 flap like [] if there is a sufficient reduction in the duration z -25 -50 -75 -100 [pzl] of the signal. All other things being equal, the removal of a significant portion of the duration of /z/ (between 75% Figure 4. Percentages of identification as the and 100%) guarantees that there will be no perceptual corresponding token containing the flap for the pair with a difference between pairs like lizard/littered, razzle/rattle, central vowel: puzzle/puddle. loser/looter, causer/caught her and puzzle/puddle. Finally, in the case of the tokens with the central vowel Thus, these results are taken as a possible explanation for puzzle/puddle, the same general trend can be observed, the historically attested change from tongue-tip fricatives although again there are certain differences. Clearly, the to alveolar flaps known as rhotacism. A logical removal of 100% of /z/, like in most of the other interpretation of the results obtained here would be that utterances, guarantees that all speakers will perceive a the intrinsic duration of /s/ or /z/ can become reduced as a flap, that is, there is no distinction between puzzle and result of factors like syllable position, stress or speaking puddle. For the other levels, however, there seems to be rate. As a consequence of this initial reduction, the more of a gradation here than in the other cases, since articulatory requirements for the production of a canonical fricative /s/ or /z/ are not achieved. In particular, the REFERENCES necessary duration that is required for the buildup of pressure which is characteristic of the fricative is not [1] H. Hock, Principles of Historical Linguistics. The reached. The leftover vocalic transitions are then Hague: Mouton de Gruyter, 1991. reinterpreted as a flap. [2] J. Romero, "Temporal reduction effects in diachronic As mentioned in the introduction, these findings can be change: rhotacism": Paper presented at the Conference relevant for the notion of gestural integrity as understood on the Phonetics-Phonology Interface, ZAS, Berlin, in the theory of Articulatory Phonology. It is observed Germany, 2001. here that, with a comparable constriction [3] C. Browman and L. Goldstein, "Towards an location—alveolar for both /s/ and []—a change in the articulatory phonology” Phonology Yearbook, vol. 3, duration of the tongue-tip gesture associated with /z/ can pp. 219-252, 1986. trigger a change in the constriction degree of the perceived segment. While a fricative like /z/ would be specified as [4] C. Browman and L. Goldstein, "Articulatory having a critical constriction degree, that is, creating a Phonology: an overview” Phonetica, vol. 49, pp. 155- narrow enough constriction to cause turbulent airflow, a 180, 1992. flap would require a closed constriction degree, since we assume that in flaps the tongue tip actually contacts the [5] D. Honorof, Articulatory Gestures and Spanish Nasal alveolar ridge. Even though the data obtained here do not , unpublished doctoral dissertation, Yale allow us to ascertain whether flaps actually do require a University, 1999. complete closure or not, we can conclude that they do not, since reduced /z/ is identified as a flap consistently. The most plausible interpretation is that in fact the perception of flaps relies more on duration than it does on constriction degree. Thus, a variety of constriction degrees, from critical to closed can be interpreted as belonging to flaps as long as the duration is short enough. Thus, there seems to be a certain degree of perceptual integration and blending between the different dimensions of the gesture.

6. CONCLUSION

The results presented in this paper corroborate the findings of previous experiments that investigated the relationship between the historical sound change known as rhotacism and the effects of articulatory reduction along the temporal dimension. Just as it was found for data in Spanish and Catalan, two languages where rhotacism occurs as a synchronic phonological process, the results presented here for flapping in American English prove the relevance of the temporal domain and of articulatory reduction in general in historical sound change. The relevance of the results presented here are another instance that demonstrate that the mechanisms that lie behind historical sound change are not different from those that are posited to explain synchronic phonological phenomena. It is suggested also that, just as in many other instances, articulatory phenomena can be seen as being responsible for both synchronic and diachronic phonological variation.