Which vowels in english are affected by syllable stress

I NI tiate. E LA borate. SMAR ter. Ful FILL. Cre ATE. MA nager. Com PU ter. Ad MIRE. Environ MEN tal. Repeat the words again. Focus on making the stressed syllable longer, louder, and higher in pitch than the others.

Pronounce the words below, taking care to use correct stress placement. Then, listen to the recordings. Were you correct? Click to see the answers Mana GER ial. IN terview. Po SI tion. BUSS iness. Each word sublist was matched as closely as possible for 1 word frequency van Heuven et al.

Thus, guidance regarding which particular stressed vowel to exchange with a given unstressed vowel and vice versa was needed. Thus, a mis-stressed word may have zero vowel errors e. In the case of the 0 Right stress manipulation, each counterbalanced sublist manipulated only degree of stress for most words —i.

Transcriptions based on the International Phonetic Alphabet for the General American English pronunciation of all stimuli and of all derivationally related word family members modeling stress manipulations were generally obtained from the Web app Lingorado Jansz, n. However, in the few cases where Lingorado failed to provide an American English IPA transcription or provided a transcription that violated the authors' American English intuitions, other online dictionaries were checked Cambridge University Press, ; Merriam-Webster, ; Oxford University Press, and standard American English IPA transcriptions were developed or revised accordingly.

This study's first author then used Ittiam Systems' free ClearRecord Lite iPhone app to record all stimuli in both their standard stress and manipulated stress forms within one of the following four neutral sentence carrier sentences:. Recording stimuli in such neutral recording frames avoided effects from either discourse-level rising intonation signaling the list of words being recorded was not yet finished or falling intonation signaling the last word in the list was now being spoken.

Stimuli were recorded within their respective carrier sentences with a slight pause before and after each stimulus word, so it could be excised from the recording without contamination from the preceding or following context. Each pronunciation was then evaluated by this study's first author and, upon her initial approval, by the second author, based on their substantial background in phonetics and phonology. Each pronunciation was evaluated within the context of its particular standard or non-standard stress stimulus set for 1 whether it clearly instantiated the target word stress manipulation, 2 whether it included all segmentals appropriately and clearly pronounced and 3 whether it exhibited comparable suprasegmental markers of stress, speed of speaking, etc.

Often, stimuli were recorded multiple times before they were deemed satisfactory. Participants were orally introduced to the experimental procedure approved by our university's Institutional Review Board and provided informed consent. Within a comfortable private cubicle, each was interviewed using an extensive Language Background Questionnaire addressing questions about their child and teenage language experience, about their English-language-learning experience and current daily English usage and proficiency, and about any L3 or L4 languages, etc.

Each trial included the following steps. The study's counterbalancing involved each L1 and L2 participant listening, in random order, to all of the Appendix 1's set A words spoken with standard stress intermixed with all set B words spoken with manipulated stress, or vice versa Appendix 2 has the words with their phonetic transcriptions.

Our word identification task used typed spellings rather than spoken accuracy as a proxy for word identification because of concerns that, particularly with standard pronunciations, it would otherwise have been impossible to identify whether participants' articulations were grounded in their having successfully identified the intended word or were instead the effect of priming leading to their likely accidentally simply repeating what they had heard cf.

One common challenge faced in studies of L1 and L2 language users Whelan, is that L1 participants generally perform relatively homogeneously, whereas L2 performance is characteristically much more variable. The current study was no exception though both groups included outliers. An additional source of variability was the wide-ranging difference in performance found across Hierarchy categories, with both L1 and L2 listeners performing for some Hierarchy categories at ceiling and for one category basically at floor.

Although several transformations i. An additional issue with non-linear data transformation is that while it can address questions of rank order, it cannot resolve questions about relative degree of impact Whelan, ; Lo and Andrews, since, for example, the square root of 25 is 5, of 16 is 4, and of 9 is 3 i. Details of all non-linear data transformations attempted are available from the dissertation of this study's first author Richards, Because this study's research questions are not so much about how L1 listeners and L2 listeners perform in relation to each other , but rather about how each group's performance compares to the predictions of our hypothesized English Word Stress Error Gravity Hierarchy, the current paper reports ANOVA analysis of the untransformed L1 and L2 listener groups' data separately.

First, we report the results for Lexical Decision LD accuracy and reaction time in light of hierarchy predictions. These two variables, respectively, measure how accurate listeners were in determining whether words were correctly or incorrectly pronounced and how long it took them to decide. Next, we report the results of the Word Identification WI task, in which listeners typed out the word they heard. This task was our proxy measure for the intelligibility of mis pronounced words across the hierarchy.

For each of this section's three parts, we present the L1 results, the L2 results, and then compare the L1 and L2 listeners. Finally, we look at how our study connects with the few others that have noted that listeners' word stress error processing appears to be predicted not only by the presence or absence of vowel errors, but also by direction of stress shift Cutler and Clifton, ; Field, The Hierarchy predicts that L1 and L2 English listeners' LD accuracy with the non-standard stress categories relatively close to standard stress will be poor but will progressively improve the further a non-standard stress pronunciation falls from the standard stress category of the Hierarchy.

Specifically, it predicts that listeners' LD accuracy will be better for words at the two ends of the hierarchy i. The L1 English listeners' LD accuracy data follow the expected pattern. In contrast, their LD accuracy with the middle-of-the-Hierarchy non-standard stress categories was poor. Table 2. Percentage of mean difference between pairs of hierarchy categories in L1 listeners' LD accuracy. Seven other L1 English listeners rated only one 0 Left non-standard pronunciation as non-standard and therefore had only one RT associated with an accurate LD for the 0 Left category.

Therefore, LD RT data across all categories of the Hierarchy was available for submission to statistical analysis for only 24 of our 38 L1 English listeners.

Figure 1. However, there is one telling exception to this overall RT trend. In other words, not only was the L1 English listeners' LD accuracy extremely low in recognizing 0 Left mis-stressings as non-standard, but on the rare occasions when they did succeed, the price tag was prolonged mental debate. These findings are not surprising since previous research has made it clear L1 English listeners have difficulty utilizing the suprasegmental word stress cues of duration, pitch and intensity that are often redundant with the more salient vowel quality cue Cooper et al.

After all, this study's 0 Left and 0 Right Hierarchy-defined non-standard stress pronunciations offered only these suprasegmental word stress cues. It is also no surprise the L1 English listeners struggled particularly to identify 0 Left word stress shifts as non-standard since, as mentioned earlier, English regularly licenses leftward stress shift for the purpose of discourse-level contrastive stress Field, Post-hoc analysis by reverse-coding L1 listeners' Hierarchy-defined in accurate LDs for the 0 Left and 0 Right categories as accurate and their Hierarchy-defined accurate LDs for these two suprasegmentally-demarcated categories as in accurate allowed us to model this question.

Thus, L1 listeners' sensitivity to the suprasegmental correlates of English lexical stress depended on the direction of stress shift. The L2 listeners' LD accuracy data visually follow a similar pattern to that of the L1 listeners Figure 2 , though apparently from a lower baseline and, as is frequently the case in studies involving L1 and L2 language users Whelan, , with much greater variability. Figure 2. Bonferroni-corrected pairwise comparisons, displayed in Table 3 , show this large effect size is due to L2 listeners' performance with the standard stress and 0 Left Hierarchy categories being reliably different from all other Hierarchy categories and the 1 Left category reliably different from all other categories except the immediately adjacent categories 0 Right and 1 Right.

Table 3. Percentage of mean difference between pairs of hierarchy categories in L2 listeners' LD accuracy. It thus visually appears Figure 3 that in cases when L2 listeners were able to make a Hierarchy-defined accurate LD with the 0 Left pronunciations, they paid an RT cost to do so. Figure 3. While the visual similarity in L1 and L2 listeners' LD accuracy data seen in Figure 2 —and even more unmistakably in Figure 4 —is intriguing, as mentioned earlier, it was impossible to test the significance of this potential LD accuracy difference because the L1 vs.

Figure 4. What should be noted from Figure 2 about L1 and L2 listeners' LD accuracy, however, is that L2 listeners' interquartile range barely overlaps with that of L1 English listeners for all non-standard stress categories in which an English word stress error induces one or more concomitant vowel errors. In other words, the L2 English listeners did not merely follow L1 English listeners' performance from a lower baseline.

Rather, the further an English word stress error fell from the standard stress category of the Hierarchy, the more L2 listeners' LD accuracy was hurt in comparison to that of L1 listeners.

Also, one additional point for future research should be noted. For most Hierarchy categories, as one might expect, L2 listeners' mean and median LD accuracy is lower than that of L1 listeners. However, the L2 listeners' mean LD accuracy in Figure 4 for the 0 Left and 0 Right categories almost exactly mirrors that of L1 listeners—and L2 listeners' median LD accuracy in Figure 5 actually exceeds that of L1 listeners.

How can this be? Figure 5. Largely in accord with L2 listeners' self-assessed English listening and speaking proficiency Table 4 , the pitch-contrastive L1 listeners' LD accuracy appears generally lower across Hierarchy categories than that of the non-pitch-contrastive L1 listeners—a finding one of our reviewers has suggested may be due to English including several short lax vowels that are not part of many East Asian languages' vowel inventory, making it difficult for speakers of these languages to accurately determine whether English words containing these short vowels have or have not been correctly pronounced.

Table 4. Specifically, the only two English Word Stress Error Gravity Hierarchy categories where the tonal or pitch-accent L1 listeners apparently outperform not only their non-tonal, non-pitch-accent L1 peers Figure 6 , but also L1 English listeners Figure 5 are the two categories where only the suprasegmental cues to non-standard stress—including the pitch cue—were available. In other words, as is characteristic of L2 speech processing generally Cutler, , retaining their L1 speech processing strategy of closely attending to the pitch cue apparently served pitch-contrastive L1 listeners well for these two Hierarchy categories.

While this study's small sample size for pitch-contrastive vs. Figure 6. The English Word Stress Error Gravity Hierarchy predicts that L1 and L2 English listeners should generally be able to recognize a speaker's intended word for English words pronounced with standard stress or pronounced with non-standard stress that is marked only suprasegmentally Bond and Small, ; Cutler and Clifton, ; Cutler, ; Small et al.

However, the further a non-standard stress pronunciation falls from the standard stress category of the Hierarchy, the more intelligibility is expected to decrease and therefore the less accurate word identification WI accuracy should become.

Inaccurate WI was defined in this study as either 1 not attempting at all to spell a speaker's intended word or 2 spelling a real English word other than what the speaker intended. Typos consisting only of added non-alphabetic characters were counted as instances of accurate WI.

Other misspellings were deleted from the data prior to WI accuracy analysis, since it was oftentimes impossible to decide objectively whether they represented 1 listeners' misspelling of the speaker's intended word that they had in fact accurately identified or 2 listeners' attempt to spell phonetically what they had heard, a response likely on at least some occasions because most non-standard pronunciations had been modeled on the standard pronunciation of a derivationally-related word family member and therefore likely sounded somewhat familiar to listeners.

After all, while some misspellings appeared to be minor misspellings, e. Table 5. Percent of WI response tokens submitted vs. We observed deterioration in L1 listeners' WI accuracy, our intelligibility proxy, with the non-standard stress categories furthest from standard stress Figure 7 left.

It is only at the 1 Right non-standard stress category that L1 listeners began exhibiting significant deterioriation in WI accuracy, either not attempting at all to spell the speaker's intended word or spelling a real English word other than that which the speaker intended. Figure 7. Like for L1 listeners, the 1 Right non-standard stress Hierarchy category is where the L2 listeners began to exhibit significant deterioration in WI accuracy—though unlike the L1 listeners, this was true for L2 listeners relative only to L2 standard and 0 Left category performance.

Both L1 and L2 English listeners experienced the greatest deterioration in intelligibility with the Hierarchy categories farthest from standard stress. However, while the visual similarity in L1 and L2 listeners' WI accuracy data seen in Figure 7 is interesting, it was again impossible to test the significance of this potential difference because of how the L1 vs. Although several i. Unattempted spellings may be the clearest possible indicator of unintelligibility as they occur only when listeners, despite being assigned no penalty for guessing, were nevertheless unwilling or unable to attempt identifying the speaker's intended word.

For both L1 and L2 listeners, the number of unique words types and percent of total words tokens they declined to identify sharply increases at the 1 Right category Table 5. However, the L2 listeners experienced substantially reduced intelligibility relative not only to L1 listeners, but even to their own standard stress performance.

The L2 listeners therefore were not performing merely from a lower baseline than their L1 listener counterparts, but rather were impacted to an even greater degree cf. The aim of the English Word Stress Error Gravity Hierarchy is to provide a means of predicting how listeners are likely to process any given word stress error. Many studies have noted the impact of vowel quality on L1 and L2 English listeners' word stress error processing Bond and Small, ; Cutler and Clifton, ; Cutler, ; Small et al.

This study adds to this research as follows. In terms of our L1 listeners' LD accuracy data, within-within-subjects ANOVA with a Greenhouse-Geisser correction found a significant interaction between the number of vowel errors and direction of stress shift, F 1. Specifically, although both number of vowel errors and direction of stress shift significantly affected L1 listeners' LD accuracy, only the number-of-vowel-errors factor did so across the entire Hierarchy.

That is, as in Field , direction of stress shift was a statistically significant factor only where non-standard stress errors were not simultaneously inducing vowel errors.

In contrast, for both Hierarchy-defined accurate and inaccurate LDs with the 0 Right stimuli, L1 English listeners' median reaction times were significantly slower relative to their accurate standard stress LDRT. In terms of L1 listeners' word identification WI accuracy, a within-within-subjects ANOVA with a Greenhouse-Geisser correction found a significant interaction between the number of vowel errors and direction of stress shift, F 1. L1 English listeners were equally accurate in identifying a speaker's intended word whether that word was pronounced with standard stress or 0 Left, 0 Right or 1 Left non-standard stress.

Thus, direction of stress shift did affect listeners' WI accuracy, but its impact was not stable across the Hierarchy.

In terms of L2 listeners' LD accuracy, a within-within-subjects ANOVA with a Greenhouse-Geisser correction found a significant interaction between the number of vowel errors and direction of stress shift, F 1. As with the L1 listeners, direction of stress shift mattered for the L2 listeners only where non-standard stress errors did not involve vowel errors.

Although only L1 listeners showed a direction-of-stress-shift-modulated effect in terms of accurate LDRT, the L2 listeners did show a direction-of-stress-shift-modulated effect when in accurately labeling 0 Left vs.

In terms of L2 listeners' WI accuracy, a within-within-subjects ANOVA with a Greenhouse-Geisser correction found no significant interaction between the number of vowel errors and direction of stress shift, F 1. In regard to the respective main effects of these two factors, however, both were significant. Unsurprisingly, number of vowel errors had the greatest impact, F 1. Nevertheless, direction of stress shift fell just shy of the 0. Specifically, L2 listeners like L1 listeners were equally accurate in identifying a speaker's intended word whether that word was pronounced with standard stress or 0 Left, 0 Right or 1 Left non-standard stress.

It was only at the 1 Right — 2 Left Hierarchy categories that the L2 listeners like the L1 listeners exhibited significant deterioration in WI accuracy. In other words, direction of stress shift did affect both listener groups' WI accuracy, but its impact was not stable across the Hierarchy.

In sum, both number of vowel errors and direction of stress shift impacted L1 and L2 English listeners' English word stress error processing. The impact of number of vowel errors and direction of stress shift, however, varied across both Hierarchy categories and dependent variables. The Hierarchy should therefore prove a useful tool for L2 pronunciation teaching and testing, as it provides an easy way of assessing the likely error gravity on any given word stress error.

Both L1 and L2 English listeners' word stress error processing largely followed the proposed English Word Stress Error Gravity Hierarchy, with stronger influence from the numbers of vowel changes and weaker influence from the direction of stress shift. However, the L2 English listeners in this study did not follow L1 listeners' performance merely from a lower baseline.

Rather, the further an English word stress error fell from the standard stress Hierarchy category, inducing one or more concomitant vowel errors, the more L2 listeners' auditory LD accuracy was hurt relative to L1 listeners. In other words, even in a non-discourse context, the further a word stress error falls from the English Word Stress Error Gravity Hierarchy's standard stress category, the more likely both L1 and L2 English listeners are to mis-segment the speech string, be led down a garden path forcing additional rounds of mental lexicon lookup, with the result of at least slowed processing reduced comprehensibility and perhaps failure to recover the speaker's intended word at all unintelligibility Bond, , ; Bond and Small, ; Cutler, , ; Isaacs and Trofimovich, It is true listeners may be able to use context to identify a mispronounced or otherwise unfamiliar word.

However, L2 listeners face an uphill battle in taking advantage of context for many reasons. This impacts L2 listeners in that their default misapplication of L1 speech segmentation strategies to the L2 stream of speech characteristically renders slow and sometimes completely unsuccessful the word boundary identification that necessarily precedes mental lexicon lookup that necessarily precedes the context-building required for recovering the meaning of mispronounced words!

In addition, context is not always particularly helpful for L2 listeners due to their less robust vocabulary and much stronger tendency than L1 speakers to hear phantom words rather than real words Broersma and Cutler, Syntactic complexity and cultural unfamiliarity can further exacerbate L2 listeners' difficulty in identifying mispronounced words from context.

These issues are compounded when the input contains multiple unrecognized forms, as less understood context from which the meaning of unknown forms can be inferred can make guessing from context untenably demanding for not only L2 listeners, but also L1 listeners Schmitt, ; Nation, ; Folse, ; Field, In sum, this study has replicated the findings of Cooper et al.

In the current study, both L1 and L2 English listeners from non-tonal L1s largely processed the suprasegmental correlates of English lexical stress merely as phonetic detail, i. The only correlate of lexical stress that these listeners consistently used to distinguish differences in word stress was not suprasegmental, but segmental.

Our phonological understanding of English word stress errors—and our pedagogy—must therefore recognize that the non-canonical use of the suprasegmental correlates of English lexical stress is generally processed as acceptable allophonic variation by both L1 and L2 English listeners.

It is only the vowel quality correlate of English lexical stress that is consistently processed categorically. Therefore, the traditional labeling of any non-canonical shift in English word stress as representing an error, regardless of whether the stress shift induces a vowel quality change, is problematic.

Traditionally, any non-canonical shift in English word stress has been treated as an error, without regard to whether the mis-stressing creates a change in vowel quality. However, L1 and L2 English listeners both frequently failed to recognize mis-stressings with no vowel errors as being non-standard.

According to the WI accuracy results, neither L1 nor L2 English listeners show any deterioration in intelligibility with 0 Left and 0 Right non-canonical stress pronunciations. Instead, for these advanced L1 and L2 English listeners, non-standard English word stress was largely defined by the presence or absence of vowel errors cf. In the Hierarchy categories farthest from standard stress, word stress errors could and did induce vowel errors that significantly reduced intelligibility Munro and Derwing, A motivation for this study was to address the importance of word stress in L2 English learning and teaching.

There is evidence that the accurate production of English word stress is crucial for intelligibility and comprehensibility.

L1 English listeners use stressed syllables to identify the beginnings of words in speech Cutler and Norris, ; Cutler and Butterfield, , and incorrect word stress can cause listeners to completely misunderstand intended words Benrabah, ; Zielinski, Incorrect word stress has also been found to lead to loss of comprehensibility Slowiaczek, ; Isaacs and Trofimovich, Because of the critical role of vowel quality in determining stress, stress errors that change vowel quality result in both L1 and L2 English listeners struggling to identify words being spoken Field, Unfortunately, due to the preference of English for alternating stressed and unstressed syllables Liberman and Prince, , non-standard word stress commonly triggers multiple vowel quality errors because stress exchange causes ordinarily reduced vowels to become clear, while adjacent ordinarily clear vowels become reduced.

Yet it is also the case that word stress errors do not invariably harm intelligibility and comprehensibility Levis, Similarly, when stress errors do not involve changes in vowel quality e. Field and this study have demonstrated that L2 English listeners are similarly affected by misplaced word stress but from a lower baseline due to causes such as the continued use of L1 processing strategies inefficient for English, phantom word activation, and a more limited vocabulary.

As a result, teaching English word stress to L2 learners is at the same time both critical and unimportant. Word stress is critical to intelligibility and comprehensibility in words where stress errors result in a change of vowel quality, with the tendency toward multiple changes in vowel quality likely to have greater impact than a single change. But there have also been influential arguments against teaching word stress that must be addressed. Jenkins' argument involves a number of implicit, problematic claims.

First, Jenkins , Jenkins says that word stress patterns vary considerably, which in light of findings about the importance of word stress for L1 English listeners would suggest there is marginal mutual intelligibility across varieties of English.

This is clearly not the case. Berg says that about 1. Second, Jenkins , Jenkins implies that if L1 speakers of English vary considerably yet understand each other, L2 speakers will also be understood despite variations in word stress patterns. Jenkins , p. As a result, materials developers and teachers of need to know specific underlying word stress patterns which facilitate perception and production of standard English word stress patterns for known and novel words Nation, ; Aitchison, ; Cutler, If these word stress regularities are learned, only the small number of relevant exceptions need to be learned individually—a much more manageable task.

Strategies such as condensing L2 learners' exposure to the similar-sounding words from which L1 listeners have acquired their implicit knowledge of English word stress patterns via rhyme-based e.

Key academic word list words such as ANalyze, aNAlysis , and anaLYtical carry a high semantic load in academic and professional communication. Unfortunately, these are precisely the type of words that are most likely to result in vowel changes when they are mis-stressed, resulting in slowed understanding at best and loss of understanding at worst.

Whether the speaker is a graduate student, business executive, healthcare worker, or one of many others whose job depends on clear communication, accurate stress of essential vocabulary can make speech more intelligible, especially when the words are longer than two syllables. This study, in order to examine the effects of word stress errors across the Hierarchy, admittedly used a task that is unlikely to show up in normal communication.

As such, it is uncertain how well the results reflect how listeners respond to word stress errors in communicative contexts. The study also is limited in its use of L2 listeners. Although there were enough L2 listeners to statistically test the research questions, the listeners came from a wide variety of L1 backgrounds, and as a result, it is unclear whether the wide variation in L2 scores came from variations in the L1s of the listeners or other unexplored factors.

The study raises questions about how intelligibility and comprehensibility would be affected in more authentic communicative contexts. Our results especially showed that greater numbers of vowel changes led to less efficient understanding and processing by both L1 and L2 listeners.

This suggests a correlation between spoken test scores and word stress accuracy, and examinations of spoken language test scores for populations such as international teaching assistants, who frequently use words from the academic wordlist Coxhead, may show that types of word stress errors correlate with test scores. Additionally, we assumed that typical word stress errors happened due to analogy with other related words Guion et al. A corpus of L2 speech that elicited varied multi-syllabic words would be useful in identifying the extent to which such analogical errors occur.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. The studies involving human participants were reviewed and approved by Institutional Review Board, Iowa State University. This article is based on the dissertation research of MG. As such, her work forms the foundation of the research. JL is her dissertation supervisor.

In this paper, he contributed most heavily to reformulating the introduction, literature review, discussion, and teaching implications. All authors contributed to the article and approved the submitted version. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Aitchison, J. Google Scholar. Anderson-Hsieh, J. The relationship between native speaker judgments of nonnative pronunciation and deviance in segmentals, prosody, and syllable structure. Balota, D.

The English lexicon project. Methods 39, — Barca, L. Word naming times and psycholinguistic norms for Italian nouns. Methods Instrum. Benrabah, M. Word stress - a source of unintelligibility in English. IRAL Int. Berg, T. Stress variation in British and American English. World Englishes 18, — Bond, Z. Listening to elliptic speech: pay attention to stressed vowels. Voicing, vowel, and stress mispronunciations in continuous speech. Broersma, M.

Phantom word activation in L2. System 36, 22— Brown, G. Listening to Spoken English, 2nd Edn. London: Longman. Brysbaert, M. Methods 44, — Concreteness ratings for 40, generally known English word lemmas. Methods 46, — Cambridge University Press Cambridge Dictionaries Online. Celce-Murcia, M. Chomsky, N. The Sound Pattern of English. Coltheart, M. DRC: a dual route cascaded model of visual word recognition and reading aloud.

Cooper, N. Constraints of lexical stress on lexical access in English: evidence from native and non-native listeners.

Speech 45, — Coxhead, A. A new academic word list. Cutler, A. Forbear is a homophone: lexical prosody does not constrain lexical access. Speech 29, — The syllable's role in the segmentation of stress languages. PubMed Abstract Google Scholar. Reed and J. Rhythmic cues to speech segmentation: evidence from juncture misperception.