Spring 2024

Andrés López, "Acquisition of metrical stress patterns in second language phonology"


This paper seeks to provide a thorough review of the current literature regarding the acquisition of second language (L2) stress systems within metrical phonology from different theoretical approaches. Metrical stress theory is the branch of phonology that postulates a hierarchical structure of stress, parsing segments into constituents such as syllables, feet, prosodic words, and prosodic phrases, to rhythmically organize stress placement (Hayes, 1995; Hammond, 1995). Acquisitions of metrical stress patterns in the phonology of adult L2 learners has been a relatively understudied topic, as compared to the sizable body of literature on L2 acquisition of other aspects of phonology, such as syllable structure or segmentals (Hancin-Bhatt, 1994, Eckman & Iverson, 1994). Among the various approaches to L2 phonology discussed in this paper, emphasis has been placed on describing selected studies which utilize parametric and constraint-based approaches to acquisition, namely Parameter Theory, a variation of the Principles and Parameters framework (Chomsky, 1981; Dresher & Kaye, 1990), and Optimality Theory (Prince & Smolensky, 1993; Kager, 1999). Various theoretical processes by which learners alter their phonological systems are described, and how factors such as age, transfer from the L1, and markedness influence L2 acquisition (Archibald, 1993; Tesar, 1997). 



It is agreed that achieving native-like pronunciation in a second language (L2), if learned after the critical period, is highly unlikely, although exceptions are possible (Ioup, 2008). Because of this, research conducted specifically on the acquisition of the L2 sound system seeks to answer why this barrier exists for pronunciation and not for other aspects of language such as syntax or vocabulary. It can be argued that a foreign accent becomes crystallized in adult L2 learners because pronunciation is the only facet of language that requires neuromuscular programming (Scovel, 1988). However, this does not answer why children in the later years of the critical period (ages 6-12) are able to acquire native-like pronunciations with ease as opposed to their adult counterparts, or why some learners can overcome the influence of the first language’s (L1) pronunciation and not others. The role of the learner’s L1 has been investigated extensively regarding L2 learnability and has been shown to be one of the main causes of a perceived foreign accent in L2 speech.  

          The focus of this review is the L2 acquisition of stress systems within the context of metrical phonology. The process by which adult learners acquire the stress systems of their target language has been a relatively understudied subject, compared to the sizable body of literature devoted to L2 acquisition of syntax or other phonological components, such as segments (Eckman & Iverson, 1994; Lombardi, 2003; Goto, 1971; Sheldon & Strange, 1982), syllable structure (Hancin-Bhatt, 1994, Hancin-Bhatt, 2000; Hancin-Bhatt & Bhatt, 1997; Broselow et al., 1998), or tone (Hao, 2012; Zhang, 2016). The acquisition of stress provides an important perspective in L2 phonology, as these features lack a uniform phonetic dimension across languages and are often not represented in the orthography. In languages such as English, word stress placement is often taught to L2 learners on a word-by-word basis, so the grammatical rules determining stress placement must be learned inadvertently. This paper reviews the current literature surrounding L2 stress acquisition and provides a summary of metrical phonology and theoretical frameworks utilized in L2 stress acquisition.  

        The primary goal underpinning the research conducted in the field of second language acquisition (SLA) is to explain how adult humans acquire a second language after having already mastered their first. Many of these studies focus on aspects of language that present difficulty to learners, motivated by the goal of explaining errors and obstacles in production and perception. For example, if a group of learners with the same L1 consistently place stress on the same incorrect syllable of word, as in qualíty or Canáda, this implies that learners’ productions are systematic and surpass simple imitation or random patterning. Second language phonology, a subfield of SLA, is the study of how adult learners acquire the sound patterns of their target languages. An important facet of L2 phonology is the use of abstract linguistic concepts, such as phonemes and allophones, to better inform our understanding of L2 acquisition based on observed behaviors in L2 learners. It has been shown through numerous studies that a learner’s L2 phonology is more than a mere imitation of the target language’s pronunciation but is in fact an independent abstract system governed by rules like those of L1 phonologies (Eckman et al., 2004). The field of L2 phonology is inherently interdisciplinary, combining aspects of theoretical phonology, psycholinguistics, and applied linguistics to adequately explain the processes by which humans acquire an L2 in adulthood. This allows findings in L2 phonology to contain theoretical implications which inform our understanding of the constraints on language learnability, an issue which lies at the core of linguistic theory. Findings in research on L2 phonology also help us to better understand the phonological and psychological processes that L2 learners undergo, allowing us to effectively formulate L2 teaching strategies accordingly. Without an empirical basis on which to center theoretical claims, the evolution of language pedagogy would remain stagnant. 

        Findings in first language acquisition (FLA) and SLA inform the theoretical bases of one another, allowing us to draw conclusions about acquisition and learnability, as well as to discover what is and is not possible regarding acquisition and linguistic universals during both stages of life. Although the body of literature centered on the acquisition of stress systems in one’s first language is much larger than that on L2 stress acquisition, this review focuses solely on acquisition patterns in adulthood, after the acquisition of the L1 has been completed. The FLA subfield of bilingual first language acquisition (BFLA) provides an important insight into how bilingualism develops in childhood and can inform theoretical approaches to bilingualism and its development later in life. BFLA studies have found that young children take about 2 years and 6 months to develop two separate stress systems for both of their languages, but interactions between these systems may occur after this age (Brulard & Carr, 2003). This implies that from an early age, humans are capable of identifying and learning two distinct stress patterns simultaneously (De Houwer, 2009, p. 173). Findings like these raise questions as to how this distinctive learning capability is maintained until adulthood, if at all. 

        Studies in L2 stress acquisition have taken a variety of theoretical and experimental approaches, all of which come together to provide diverse perspectives in the development of stress theory, acquisition theory, and learnability. Most stress acquisition studies draw heavily from metrical stress theory, with metrical approaches dominating the literature. While other studies may take an alternative approach (see Ou & Ota, 2015), the inclusion of a metrical foundation in L2 stress acquisition studies has proven effective in explaining a variety of stress pattern phenomena in L2 learners. Other non-metrical approaches have included investigations on the phonetic realization of L2 stress (Garcia & Guzzo, 2017; Nguyen & Ingram, 2005), perception of L2 stress (Dupoux et al., 2001; Peperkamp & Dupoux, 2002), and possible pedagogical implementations for teaching stress (Hayes-Harb & Hacking, 2015). Across metrical L2 stress acquisition studies, the discord among scholars lies in deciding which framework is most effective in explaining learner errors and acquisition processes. This paper provides an overview of the most prevalent stress framework in L2 stress studies, Parameter Theory (Chomsky 1981a, 1981b; Dresher & Kaye, 1999), as well as a lesser used framework which has received much more attention in other areas of L2 phonology, Optimality Theory (Prince & Smolensky, 1993; Kager, 1999; Tesar, 1997). Although these studies differ in the methods used to analyze their data, the shared primary objectives are to identify barriers in L2 acquisition, explain why they exist, and how they can be overcome. 

        The remainder of this paper is organized as follows: section 2 provides an overview of stress and metrical phonology; section 3 explains the various theoretical approaches to L2 phonology and how these approaches relate to L2 stress acquisition; section 4 details two important frameworks used in L2 stress acquisition studies: Parameter Theory and Optimality Theory; section 5 offers a survey of notable studies utilizing both Parametric and Optimality Theoretic approaches; and section 6 concludes. 


Stress Patterns and Systems 

Simply defined, stress is the relative prominence of one syllable (σ) over another. In any given utterance (whether it be a word, phrase, sentence, etc.), one syllable will bear the highest prominence, or primary stress, a universal characteristic referred to as culminativity. This manifestation of stress, or prominence, is used in the linguistic system for phonemic distinction, marking of morphological or phrasal boundaries, and emphasis (De Houwer, 2009). In English, for example, stress can be used to distinguish nouns/verbs from their denominal/deverbal counterparts1


Even in languages where stress is not phonemic or arguably present, such as French or Turkish (Özçelik, 2017), prominence tends to surface on at least one syllable due to culminativity, as mentioned above (Hayes, 1995, p. 29). In such cases, prominence can be realized at the phrasal level, rather than at the word level. This gives an argument to the hierarchical structure of stress related phenomena, which is expanded upon in metrical phonology.  

        Cross-linguistically, stress or prominence can be realized and analyzed acoustically in a variety of ways, such as through pitch, intensity, or duration (Garcia & Guzzo, 2017). The problem therein lies in developing a framework to which one can accurately apply and examine all existing observations of stress patterning. Any attempt at a theoretical framework that describes stress patterns must be able to accommodate all variations of stress realizations, regardless of how they are realized phonetically. As best described by Hayes (1995), stress in and of itself is not defined by its acoustic features but more so by its relative prominence to other syllables in a given utterance. Because of this, a non-linear structural framework is necessary to accurately analyze stress patterns with respect to their relativity and not their phonetic properties. The central focus of the current paper is the abstract representations of stress in language users (as opposed to their phonetic realizations) and the processes by which stress is realized and learned. 

        There are a variety of ways that languages can assign stress to words or phrases. In some languages, stress is predictable, or fixed, meaning one can deduce the correct stress placement on a new word without having heard it before. For example, in Finnish, primary stress always falls on the initial syllable of a word (Carlson, 1978): 

This notion of predictability can be defined by phonological factors, such as syllable weight or syllable position. This type of stress system is often labeled as grammatical, as stress placement is deduced by grammatical rule. In other languages, stress can be unpredictable, or free, meaning that the primary stress placement in any given morpheme must be memorized and learned with each new word. The line between free and fixed stress systems is not clear cut, with most languages incorporating aspects of both types of systems. In languages with lexical stress, stress is said to be encoded lexically, or ‘stored’ in the lexicon along with the word’s segmental structure. This stands in contrast to linguistic structures which are computed, e.g. syntactic structures such as sentences or phrases (Archibald, 1994). In Modern Greek, for example, primary stress in nouns is lexically encoded and unpredictable by phonological rule. Although stress placement is unpredictable, it is limited or “fixed” to the final three syllables of a word, and secondary stress can be deduced grammatically based on the placement of primary stress (Hayes, 1995, p. 204). Thus, Modern Greek’s stress system can be said to contain aspects of both fixed and free stress systems. 

        Given the diverse typology of stress systems in languages across the world, contact between these systems in the interlanguage of an L2 learner can pose difficulty in acquiring the new metrical system of any given language. For example, when learning a language with bounded stress rules2, speakers of languages where stress is unbounded must first discern where these metrical boundaries are using only overt forms as input, and then apply these boundaries to their running speech, while also retaining these rules for productive application when faced with new vocabulary. Many of the studies in L2 stress acquisition investigate the learnability of rule-based or grammatical systems, as this steps away from the need to lexically encode stress placement on an item-by-item basis. This allows for a thorough analysis of the process that learners undergo, a process which would otherwise depend largely on memorization. 


Metrical Phonology 

        The most prominent phonological framework in which stress patterns are analyzed is Metrical Stress Theory, or Metrical Phonology (Hayes, 1981, 1995; Hammond, 1995). Metrical phonology postulates that certain syllables in an utterance receive stress or prominence based on an abstract hierarchical system of prosodic phonological units that organizes syllables into a rhythmic structure. These prosodic constituents are purely abstract with little to no phonetic realizations, and their existence is deduced through thorough analysis of stress placement patterns across languages. The metrical approach to stress patterns abandons the notion of stress as a feature, representing stress non-linearly as a linguistic manifestation of rhythmic organization. Metrical phonology is used to explain the observed regularities of cross-linguistic stress systems, whether at the word level, the phrase level, or the utterance level. 

        One of the main tenets of metrical phonology is the foot (Ft, less commonly Σ), which is the minimum constituent used when grouping stress patterns or assigning stress rules (Hayes, 1995, p. 40; Roca, 1992; Selkirk, 1980). Feet can either be disyllabic, made up of two syllables, or monosyllabic, made up of one heavy syllable. This depends on syllable weight and foot binarity minimums and maximums in a language, which will be discussed shortly. Within a foot, one syllable (the head syllable) carries more prominence over the other. Left-headed feet (σ́ σ) are considered trochaic, while right-headed feet (σ σ́) are considered iambic. In multisyllabic words, syllables are grouped into feet, which are subsequently grouped into larger constituents, such as the prosodic word (PWd/ω) or the prosodic phrase (PPh/φ), thus creating a hierarchical structure. Feet can be built beginning at the right edge of the word and moving towards the left, or from left to right. For example, the stress assignment rule for English nouns containing more than three syllables is as follows: skip over the final syllable and build trochaic feet from right to left, with main stress falling on the head syllable of the rightmost foot (Hammond, 1995, p. 318). Examples of unparsed stress patterning in English are given below in (3): 

Various methods for representing such groupings have been proposed, such as tree theory (Liberman & Prince, 1977), or the bracketed grid (Halle & Vergnaud, 1987; Hayes, 1981). The representation adopted here is the version of the bracketed grid seen below in (4) introduced by Hayes (1995, p. 39), where an ‘x’ represents a stress-bearing syllable and ‘.’ represents a stressless syllable. The bottom row represents the foot level, where foot boundaries are marked by parentheses. The second row represents stress at the PWd level, with edges marked by parentheses, and the third row represents stress at the level of the PPh. Metrical groupings of various English words are shown in (4): 

        The final syllables of the words ‘municipality,’ ‘animal,’ and ‘football’ are notably unfooted, despite being at the right edge of the word. This phenomenon is known as extrametricality, a process by which final syllables of a word are rendered invisible when parsing a syllable metrically or applying stress placement rules (Hayes, 1995, p. 57). In English and other languages with extrametricality, such as Spanish, the extrametricality of a syllable or final consonant is lexically marked, thus requiring memorization, or lexical encoding (Harris, 1983). The type of extrametricality used within a language can change depending on word class. In English, extrametricality in nouns typically affects the final syllable, while in verbs, extrametricality only affects the final consonant3. This difference is displayed in (5), using words with identical syllable structure (VC.CVCC). 

Syllable Weight 

        Syllable weight, or syllable quantity, plays an important role in stress assignment among languages and will be explained here briefly. In moraic theory (Hyman, 1984; McCarthy & Prince, 1986; Hayes, 1989), syllable quantity is defined in terms of the mora (μ) which acts as a sub-syllabic abstract unit of weight. A light syllable is said to have one mora, whereas a heavy syllable is said to have two. Morae have no phonetic representations, although they can be associated with surface level phenomena such as vowel length. In other words, they are a purely abstract concept, and their existence must be inferred through the observed behavior of syllables in phonological processes. 

        Within a syllable’s geometry, there is the onset and the rime, with the onset consisting of any segments before the syllable’s most sonorous peak and the rime consisting of all else (Hayes, 1981, p. 25). The rime can be further divided into the nucleus and the coda, the former of which is usually a vowel (or syllabic sonorant) and the latter usually being a consonant. The distribution of mora bearing segments differs depending on the language. Some languages consider vowels to be the only moraic segment, such as in Yidiɲ. Other languages consider both vowels and sonorants to be moraic, as in Lithuanian, while languages like English, Cairene Arabic, and Latin treat vowels, sonorants, and obstruents as moraic (Zec, 1995, p. 89). Universally, CV syllables are considered light, as in (6a), and CVV syllables are considered heavy, as in (6b). As stated above, CVC syllables can be considered either heavy (6c) or light (6d) depending on the language. 

        Languages that factor in syllable weight during stress assignment are considered quantity sensitive (QS), in that stress placement is sensitive to the weight of a syllable. The opposite systems, in which syllable weight is ignored, are referred to as quantity insensitive (QI). Foot binarity minimums in QS languages may allow for bimoraic feet made up of a single syllable, instead of strictly disyllabic feet. Such is the case in English, where bimoraic long vowels and closed syllables can make up a single foot, as seen in (7). In QS languages, heavy syllables will attract main stress over light syllables, a concept known as the Weight-to-Stress Principle (WSP; Prince, 1990). In English, a QS language that adheres to the WSP, most nouns are stressed on the antepenultimate syllable of a trisyllabic word, unless the penultimate syllable is heavy, as exemplified below: 

Theoretical Approaches to L2 Phonology 

Since its inception as a subfield in the 1950s, one of the most widely accepted generalizations in L2 phonology has been the undeniable role that the L1 plays in acquiring an L2’s sound system, or transfer. Early approaches to L2 phonology sought to explain ease and difficulty of acquisition through comparisons of phonologies’ segmental inventories. This gave rise to the Contrastive Analysis Hypothesis (CAH) which argued that L2 speech patterns can be explained by L1 transfer and differences between the L1 and L2 (Lado, 1957). Under this framework, asymmetries in the phonemic and allophonic distribution of phones in a language’s phonology cause the most difficulty for a learner. For example, an English speaker acquiring an Arabic sound system would need to learn to distinguish the three voiceless fricatives /h/, /x/, and /ħ/, all of which are separate phonemes in various Arabic dialects. Because English’s phonology only contains one of these phonemes, /h/, an English speaker may have difficulty perceiving and producing the contrast in Arabic. For differences in allophonic distributions, an American English speaker who considers the phones /t/ and /ɾ/ allophones of the same phoneme may have difficulty acquiring Spanish phonology which considers these two separate phonemes. While the factors mentioned above undoubtedly play a role in L2 phonological acquisition, the framework of the CAH has been abandoned in favor of other approaches. One of the main weaknesses of a comparative approach is the undefined criteria for determining similarity among speech sounds, a question which remains unresolved up until today (Major, 2008). Another weakness is its inability to explain patterns or substitutions which cannot be attributed to either the L1 or the L2, a phenomenon usually attributed to the effects of universal constraints on linguistic structure. In terms of L2 stress acquisition, purely comparative approaches may attribute learning difficulties to differences in typological properties of the two stress systems. For example, a speaker of an L1 in which stress is non-contrastive may have difficulty perceiving and producing lexically contrastive stress patterns in an L2. However, as with segmental analyses, comparative approaches for suprasegmental acquisition fail to describe the mechanisms and processes that learners undergo, as well as variation among learners of different experience levels. 

        The role of perception in acquiring phonemic contrasts has been investigated extensively based on the assumption that one must be able to distinguish a contrast in perception before being able to produce such a contrast systematically, as is the case with children and FLA (Smith, 1973). Certain studies have shown, however, that adult learners are capable of producing segmental contrasts that they cannot perceive (Goto, 1971; Sheldon & Strange, 1982), a finding that Eckman (2004, p. 519) ascribes to the difference in literacy between FLA and SLA; children acquiring an L1 are preliterate, as opposed to the literate adult learners participating in these studies. Dupoux et al. (2001) found that L1 French learners of Spanish had difficulty in perceiving lexical stress contrasts in pseudowords, thus pointing towards a construct of stress deafness in L1 speakers of stressless languages. A later study by Peperkamp & Dupoux (2002) presented the same findings of stress deafness in speakers of Finnish and Hungarian, two languages with fixed, non-contrastive stress. However, this same study found the opposite for speakers of Polish, with Polish speakers maintaining the ability to perceive and distinguish between minimal stress pairs. As will be discussed later, Polish stress is thought to be fixed on the penultimate syllable, with antepenultimate and final stress occurring in exceptional loanwords. The authors ascribed Polish speakers’ ability to perceive stress contrasts to these exceptions, stating that Polish speakers must retain stress in their phonological representations as adults, unlike speakers of French, Hungarian, and Finnish, in which no alterations in stress placement are present. The model proposed here, later termed the Stress Deafness Model, claims that the higher the predictability of stress in the L1, the harder it will be for a native speaker to perceive stress contrasts in an L2. 

       Several other models have been proposed in attempts to understand the role of perception in the production of L2 contrasts; namely the Perceptual Assimilation Model (PAM; Best, 1993, 1994, 1995), the Feature Competition Model (Hancin-Bhatt, 1994), and the Speech Learning Model (SLM; Flege, 1995). Each of these models seeks to explain how speakers perceive L2 segmental contrasts not present in the L1 and how they subsequently categorize these sounds based on their features or similarity to L1 sounds. Among the similar claims of the SLM and PAM, the main conclusion is that L2 speech sounds that are phonetically similar to L1 speech sounds will be more difficult to produce and perceive due to assimilative categorization. L2 speech sounds that are uncategorizable will be placed into newly created phonetic categories, thus making them easier to perceive and produce. As demonstrated by Flege (1987), advanced L1 English learners of French were able to produce the French high front vowel /y/ authentically, whereas the high back vowel /u/, which has a rough counterpart in English, was substituted with an English-like pronunciation. The dissimilarity of French /y/ from any other sounds in English thus required a new phonetic categorization, whereas French /u/ was assimilated with English /u/, thus halting further learning. These models can be modified and applied to the acquisition of non-categorical suprasegmental features such as tone and vowel length, which each have two separate phonetic dimensions, pitch and duration, respectively. The phonetic dimension of stress, however, is realized differently among languages, with some utilizing duration, pitch, intensity, or a combination of the three to differentiate a stressed syllable from an unstressed one. The phonetic realization of stressed syllables is also uniform within a language, and is not used for phonemic distinction, as are contrasting tones and vowel length. These factors pose a problem for application of a model in which features are learned based on phonetic similarity.  

        In L2 learning environments where stress placement is explicitly taught, phonetic cues for stress are typically ignored. This raises the question as to how stress is represented phonetically after ultimate attainment, as it would represent a form of subconscious learning through exposure. Garcia & Guzzo’s (2017) study involved the surface representations of L2 English stress in Québec French (QF) speakers. The authors presented two main findings. First, in their L1, QF speakers demonstrated no phonetic representations of word level prominence or constituency, meaning that QF does not realize foot structure or word stress on the surface. This means that a phonetic realization of word stress is something that must be acquired by QF speakers learning English. Secondly, in their L2 English speech, QF speakers utilize duration as the main phonetic dimension when realizing English stress, demonstrating the acquisition of non-phonemic phonetic cues. Another study investigating the acquisition of phonetic cues of English stress in Vietnamese speakers presented similar findings. Nguyen & Ingram (2005) found that L1 Vietnamese learners also acquired duration as the main phonetic dimension for representing stress, along with pitch. Vietnamese is a tonal and stressless language, which uses pitch to determine lexical meaning in words. Pitch was used by beginner and advanced learners alike when identifying stress placement in English words, but in production, the advanced learners were able to produce native-like duration patterns in both stressed syllables and unstressed syllables. Both studies present similar findings in that duration seems to be the main phonetic representation acquired by learners, regardless of whether stress or duration is used in the L1. 

        The role of linguistic universals also plays an important role in L2 acquisition. The language faculty that all humans are born with, commonly known as UG, is hypothesized to impose universal constraints on language, allowing for variation in phonological structure across languages while also preventing it from becoming unlearnable. Identifying these universals in linguistic structure is one of the main goals of linguistics as a discipline. For example, a universal characteristic of all languages is the concept of stress culminativity mentioned previously, being that one syllable in any given prosodic constituent will always receive main prominence, regardless of whether stress is phonemic or contrastive in the language (Hayes, 1995). No language displays primary stress on more than one syllable within a PWd, demonstrating the universality of this particular linguistic constraint. The concept of an interlanguage (IL; Selinker, 1972), brought about through research on universals and L2 acquisition, views the learner’s mental state as a unique middle-grammar that is separate from that of the L1 and the L2. This intermediate stage is viewed as its own linguistic system, governed by the same universal constraints as ordinary phonologies. It is assumed that FLA is primarily governed by UG principles, and adult acquisition is thought to involve these principles to a certain extent (White, 2003). The process of proving this adult access to UG is one of the main goals of L2 stress acquisition research, which is accomplished through various approaches as will be described below. 

        Research on L2 phonology has repeatedly reported the emergence of patterns that cannot be attributed to either the phonological system of the L1 nor the L2, all while occurring in other unrelated languages, providing evidence for the role of cross-linguistic universals in L2 acquisition. Concepts such as markedness have been used in conjunction with transfer to explain such patterns. The concept of markedness assumes that certain binary classes of features (voiced vs. voiceless stops, oral vs. nasal vowels) are differentiated based on occurrences within and across languages, with more common features holding some form of distributional privilege over the other. The class of features which is much more prevalent across languages is deemed unmarked, while features which are less common cross-linguistically are considered marked (Eckman, 2008, p. 96). For example, oral vowels are present in practically every attested language, while phonemic nasal vowels are only present in a subset of the languages that contain oral vowels. This demonstrates that for whatever reason, oral vowels are preferred over nasal vowels, whether it be for their simplicity, naturalness, or ease of articulation. The Markedness Differential Hypothesis (Eckman, 1977) and the Ontogeny Phylogeny Model (Major, 2001) both implement the concepts of markedness, universals, and L1 transfer in explaining L2 acquisition. The Full Transfer Full Access Hypothesis (Schwartz & Sprouse, 1996) states that the initial state of the learner’s phonological system when acquiring an L2 is practically defined by the native language, with the learner’s interlanguage being initially similar to that of the L1. This position is widely accepted in the domain of L2 stress acquisition, as will be discussed in section 5. 

        Some approaches to L2 stress acquisition have questioned the extent to which metrical structures are involved in the process of L2 stress acquisition, especially when involving an L1 that is considered stressless, leaning more towards a model of generalization and analogy. In a study involving Mandarin L2 English learners, Ou & Outa (2015) state that the question as to whether metrical structures are involved in the acquisition of L2 English stress is inconclusive. Participants in this study were presented with English pseudowords, categorized as nouns and verbs, each consisting of a specific phonotactic structure in the final syllable, e.g. CVCC-final verbs, CV-penultimate nouns, etc. The study found that L1 Mandarin learners were able to dissociate CVCC-final verbs (bazent, zasept) from CVCC-final nouns (fermekt, shalent) by correctly placing final stress on the former (bazént, zasépt), and penultimate stress on the latter (férmekt, shálent), potentially displaying an acquired sensitivity to extrametricality differences based on word class. When presented with CV/CVC-penultimate nouns and verbs (natipa, teremy/natimpa, pemisto), however, the learners did not display any distinct stress assignment behavior, contrary to their treatment of heavy syllable final nouns and verbs. In other words, Mandarin learners learned to stress a heavy syllable if it is the last syllable of a verb, but failed to do so systematically when the heavy syllable was penultimate, yielding incorrect patterns such as natípa, terémy or nátimpa, pémisto. Ou & Ota claimed that the productive stress assignment patterns of the L1 Mandarin learners could be reduced to a simple generalization that disyllabic English nouns receive penultimate stress, and disyllabic English verbs receive ultimate stress. They support this claim by citing estimates of stress distributions: 94% of disyllabic English nouns receive penultimate stress, with 69% of disyllabic verbs receiving ultimate stress (Kelly & Bock, 1988). The authors conclude that the role of metrical structure in L2 stress acquisition is not clear when the L1 is not governed by metric principles. They also state that observed patterning of stress placement should not be considered metric in nature unless they co-occur with “clustered metrical properties” or can be explicated adequately (Ou & Ota, 2015, p. 407). As noted by Pater (1997, p. 235), learning of stress placement in a L2 could easily be ascribed to the mere memorization of stress patterns or to analogical comparison with similarly structured words. This assumption is entirely plausible and can easily explain why more experienced learners tend to assign stress correctly compared to their less experienced counterparts. However, this viewpoint fails to explain why some errors are more common than others within a group of learners. Due to this complication, numerous studies have adopted rare words or pseudowords as stimuli, to bypass any potential memorization, and focus solely on the computed placement of stress. The question of whether mechanisms such as memorization or analogy play a significant role when learning stress placement is still up for debate, as a variety of other languages use these mechanisms by means of encoding stress in the lexical description.  

        The role of orthography in the acquisition of a L2 sound system has also received some attention in the literature. Previous studies have demonstrated that orthographic input can significantly alter the way learners acquire segmental features of a language’s phonology (Bassetti, 2006; Hayes-Harb et al., 2010). In an experiment centered on English speaking learners of Russian, Hayes-Harb and Hacking (2015) investigated whether orthographically marked word stress would facilitate the stress placement of newly learned words. Both experienced and novice learners of Russian were presented with twelve pseudowords forming six minimal stress pairs along with their pronunciations and an image indicating their meanings. Control groups were presented with a standard orthographic representation (without stress markings), and experimental groups were presented with representations which included marked stress, e.g. <суба> vs. <субá>. After a short period of intensive learning, participants were then shown an image accompanied with a sound recording of one of the pseudowords and asked if the two stimuli matched. Interestingly, the authors found no improvement in word recognition accuracy across the groups, indicating that stress placement errors cannot be overcome by changes to the orthographic input. This study does not focus on the accuracy of stress production, however, which has been shown to differ from accuracy in stress perception (Archibald, 1993; Goto, 1971; Sheldon & Strange, 1982). 

In short, the role of transfer on L2 phonological acquisition has been investigated extensively, demonstrating the undeniable effect that the native language has on pronunciation. Nearly every theoretical framework in any subfield of L2 phonology acknowledges the role of transfer in one way or another, although approaches differ in how they explain deviant patterns not attributable to the L1 or the target language. 


Frameworks in Stress Acquisition 

This section provides background information regarding the two most prominent theoretical frameworks used in L2 stress acquisition studies, Parameter Theory (PT) and Optimality Theory (OT). Both frameworks acknowledge the importance of transfer in L2 phonology, as well as the role of universals in acquisition. In PT, the initial state of the IL is defined by the parameter settings of the L1, whereas in OT, these initial parameters are translated as the L1’s constraint rankings. With respect to universals, both frameworks consider their structures to be universal, with every language having the same parameters with different settings, or every constraint with different rankings. In OT, the role of markedness is explicitly represented through markedness constraints. 


Parameter Theory 

        Much of the work done on L2 stress acquisition follows the framework of Parameter Theory, a version of the Principles and Parameters framework outlined in Chomsky (1981a, 1981b). It is postulated that stress placement rules are governed by the variable settings of binary parameters, which represent certain truths within the system. For example, the parameter labeled ‘Extrametricality’ is set to ‘Yes’ in English, meaning that when stress rules are computed, extrametricality of the final syllable or consonant must be taken into account. Every language is governed by the same metrical parameters, and variation in stress rules among languages is ascribed to all possible combinations of settings for each of the parameters. In other words, any combination of parameter settings should hypothetically yield a possible stress assignment system in any natural language. The universal metrical parameters of stress systems are presented below in (8), taken from Dresher & Kaye (1990): 

        The parameter settings relevant to English stress placement will be demonstrated for explanatory purposes in (9). Each parameter will be presented with its English setting along with a subsequent step-by-step parsing of a five-syllable word made up of all light syllables (nominality), as well as a four syllable word with a heavy penult (elemental). Note that the ordering of the parameters is arbitrary. 

        Parameters are assumed to be set during acquisition of one’s first language, which are inferred based on positive and negative input. The acquisition of a second language with different parameter settings would require the resetting of these parameters, such as resetting ‘P8: Extrametricality’ from ‘Yes’ to ‘No.’ This process of resetting assumes that the initial state of the IL has the same parameter settings as the L1, instead of neutral ‘unset’ parameters. Various studies involving L2 stress acquisition provide evidence for such parameter resetting, comparing both the L1 and L2’s parameter settings to the parameter settings of the IL. Özçelik (2016, 2017) argues that the acquisition of prosodic parameters in childhood and adulthood occurs in stages, following a learning path based on a hierarchical structure of the prosodic parameters. He refers to this as the Prosodic Acquisition Path Hypothesis. The model predicts that the earlier a parameter is set on the hypothetical learning path during childhood, the harder it is to reset as an adult. He provides evidence for this hypothesis through the inability of English learners to rid their speech of the foot when reading Turkish words (a language he argues is footless4), because the parameter which determines if a language is footed or not is ranked the highest in the parameter hierarchy. Others have questioned the possibility of parameter resetting in adulthood (Stozer, 1994), with van der Pas & Zonneveld (2004) providing a comprehensive analytical critique and reinterpretation of three widely cited L2 stress acquisition studies, which will be discussed in section 5. 


Optimality Theory 

        In contrast to the rule-based parametric approach to L2 stress, Optimality Theory, developed by Prince and Smolensky (1993), provides for a constraint-based approach to L2 stress placement and learning that involves the ranking and reranking of violable constraints. OT is an input-output based framework in which potential output candidates derived from underlying representations are filtered through a set of violable constraints, after which the most optimal candidate, or the form that violates the least number of constraints, is realized on the surface. The ranking of each constraint is important, as violations of highly ranked constraints are seen as more grave than low ranking constraints. OT constraints are said to be universal, meaning that every language’s grammar is composed of the same constraints, with the only difference being their ranking. A constraint ranked highly in one language might be ranked so low in another that its effects are never considered in determining output forms. An important component of OT grammars is the power to predict all possibilities of phonological patterning through a constraint ranking’s factorial typology. In other words, all possible permutations of a given set of constraints should account for all attested structural variation in grammars, with analyses often overpredicting the range of attested patterns.  

        The constraints which dictate metrical patterning have been well developed within the OT framework. The universal constraints proposed for metrical stress are as seen in (10), from Tesar (1997, p. 3). The constraints below are for bounded stress systems and can be freely ranked. 

OT analyses are represented in a tableau, as in (11). The leftmost column contains the input form/underlying representation, along with the generated possible output candidates. Along the top row, constraints (C) are presented in descending order of rank. An asterisk (*) denotes one violation of a constraint, with (!) indicating a fatal violation, meaning that the candidate is out of the running. (☞) denotes the most optimal candidate, or the form that incurred the least number of violations. Shaded boxes are included for clarity, signifying a candidate is no longer in the running. 

        In (11), candidate 4 is eliminated first, with candidates 1 and 2 following in subsequent rounds. Candidate 3 surfaces as the most optimal, since it has incurred no violations. A simplified analysis of the French stress system will be used for demonstration. It is agreed that French is a language without lexical stress, with prominence falling on the final syllable of a prosodic phrase (Dell, 1984). The question of whether this prominence is purely intonational or a product of metrical parsing is still up for debate (for a discussion see Goad & Prévost, 2011). For the sake of simplicity, this analysis assumes that French phrase-final prominence is a representation of stress in the form of a right-aligned iambic foot. Thus, the two constraints that will be used are IAMBIC and MAIN-R. The constraints are defined below: 


         Both IAMBIC and MAIN-R are assumed to be undominated in French, since there are no occurrences of trochaic feet or left-aligned head feet. Since the grammatical candidate does not violate either constraint, no relative ranking between the two can be determined. Thus, the two constraints are demarcated by a dashed line. The remaining constraints in (10) which were not visibly represented in this analysis are hypothetically ranked below MAIN-R, although their ranking is irrelevant since they will most likely never come into play. The tableau for stress placement in the word escargot is presented in (13). 

In (13), the second and fourth candidates are immediately eliminated due to the presence of a trochaic foot. The third candidate is then eliminated since the head foot is one syllable away from the right edge of the word. This leaves the first candidate as the most optimal, incurring no violations. The first candidate’s violations of unmentioned, lower ranked constraints do not matter, since it has already been selected as optimal. There exist other factors which must be accounted for in a complete analysis of French stress, such as iterativity and weight sensitivity, although these will not be discussed here. 

        It is important to mention that in the domain of L2 stress acquisition and L2 acquisition in general, the constraint ranking of one’s native language is said to be the initial state of the IL, with learning occurring through the reranking of constraints. There have been various proposals as to how this reranking process occurs, such as the Constraint Demotion Algorithm (Tesar & Smolensky, 2000; Tesar, 1997), and the Gradual Learning Algorithm (Boersma, 1997). These systematic reranking strategies demote and promote specific constraints based on positive and negative input from overt forms, with the target language’s constraint ranking being the end goal. OT’s longstanding presence in the field of L2 phonology demonstrates the effectiveness of constraint-based analyses in explaining IL phenomena. Constraint interaction and reranking processes can provide alternative explanations for occurrences not adequately explained through a rule-based or comparative approach. Previous OT-based studies in L2 phonology have mainly focused on syllable structure (Hancin-Bhatt & Bhatt, 1997; Hancin-Bhatt, 2000; Broselow et al., 1998), with other studies investigating segmental substitution (Lombardi, 2003) or tone (Zhang, 2016). Relatively few L2 stress acquisition studies have utilized an OT framework, with most studies adopting a Parametric approach to stress acquisition. 


Studies in L2 Stress Acquisition 

This section presents the findings of various studies investigating L2 stress acquisition. Due to the nature of this paper, an exhaustive review of the current literature is not feasible. However, the studies selected will provide a diverse range of findings and demonstrate the various approaches one can take when investigating L2 stress. Although the most common finding in these studies is unsurprisingly the role of L1 transfer, the methods by which these authors explain the acquisition path of their learners is highly informative. For either framework, an older study and a newer study are presented to offer perspectives from two distinct time periods within the field. 


Parametric Approaches 

        In some of the earliest and most referenced investigations in L2 stress acquisition, Archibald (1993, 1994, 1998) presents findings that suggest the acquisition of English stress parameters by L1 Hungarian, and Polish speakers. In Polish, primary stress falls on the penultimate syllable of a word, regardless of syllable weight5 (Archibald, 1994; Hayes & Puppel, 1985). If two or more syllables precede the head syllable of a word, secondary stress will fall on the initial syllable. In other words, two syllabic trochees are aligned with the right edge and the left edge of the word, with primary stress falling on the right-aligned foot. Examples are provided in (14). 

        Hungarian is a language that has fixed main stress on the first syllable of the word, with iterative trochees built from left to right. Hungarian is QS, however like Mongolian, coda consonants are not moraic, with syllable quantity determined by vowel length only (Archibald, 1993, p. 47). Since primary stress is fixed to the initial syllable, syllable weight only plays a determinant role in the assignment of secondary stress. Archibald (1993) assumes the analysis of Hungarian secondary stress presented in Kerek (1971), with secondary stress falling on every other syllable following the head, except for the last (σ́ σ σ̀ σ σ). If the third syllable is light and the fourth syllable is heavy (Ĺ L L H̀ L L̀ L), secondary stress skips over the light syllable and falls on the heavy syllable and continues to alternate thereafter. Examples are provided in (15): 

        The variation in stress patterning between these three languages are represented by the differences in parameter settings seen below in (16). Polish learners are predicted to have difficulty acquiring quantity sensitivity and extrametricality, while Hungarian learners are expected to struggle with right-edge alignment, quantity sensitivity, and extrametricality.  

For the Polish participants, stress placement errors were usually on the penult, which would be an expected result of transfer. There were no distinctions made between light and heavy syllables, however, indicating that the QI parameter was not reset. Some instances of trisyllabic CV.CVC.CV nouns were realized with antepenultimate stress (ágenda, ároma), possibly indicating a resetting of the extrametricality parameter. These were ascribed to Polish’s restricted extrametricality, however, which could have been overextended to English words. This implies that Polish’s Extrametricality parameter was originally set to yes and did not change and that no other parameters were reset. The Hungarian participants displayed quantity sensitivity to long vowels in English words but not to heavy syllables with coda consonants. For example, the Hungarian learners would stress decíde and aróma correctly but would stress óbserve, and cóllapse on the penult. This error is described by Archibald as a generalization of L1 parameter settings regarding weight sensitivity and implies no resetting of any parameters.  

        The main conclusions drawn by Archibald from these findings are that the majority of errored stress placement is determined by transfer of L1 parameter settings, that the adult learners can reset parameters to the correct L2 settings, and that the interlanguage grammars of adult L2 learners do not violate metrical universals (Archibald, 1994, p. 230). The author also presents data that implies the resetting of certain parameters; namely quantity sensitivity and insensitivity, extrametricality, and word-tree dominance, although the specific parametric alterations are not explicitly explicated. It is important to note that the majority of participants were able to place stress on the correct syllable, with incorrect placements making up around 25% of the responses on average. Since the study’s stimuli was made up of real English words, this rate of success could be ascribed to previous knowledge of stress placements in English words, or alternatively, to parameter resetting. As mentioned above, studies conducted using pseudowords render this prior knowledge obsolete. Nevertheless, while correct stress placement could be attributed to previous experience, the common errors produced by participants could not be adequately explained through intermediate stages of parameter resetting. 

        The analysis of the data presented in these studies has also been called into question by van der Pas & Zonneveld (2004), who cite variable interpretations of the study’s findings. They mention Pater (1997) and Youssef & Mazurkewich (1998) specifically, who reached opposite conclusions when referencing Archibald’s work, with Pater (1997) stating that Archibald (1993) demonstrated successful parameter resetting, and Youssef & Mazurkewich (1998) claiming that the adult learners in Archibald’s study were not able to reset these parameters. van der Pas & Zonneveld (2004, p. 148) also argue that Archibald misinterpreted the analysis of Hungarian quantity sensitivity presented in Kerek (1971) and claim that CVV as well as CVC syllables are considered heavy, which pose further questions regarding the credibility of the author’s interpretations. These differing conclusions, along with the lack of clarity in demonstrating parameter resetting, renders Archibald’s studies as largely inconclusive regarding the parametric approach. These studies do, however, provide valuable empirical evidence which contribute to our understanding of transfer in L2 stress acquisition. 

        In a recent study from Özçelik (2021), stress placement in the production of Mongolian words by native English speakers was investigated under parameter theory. The stress system of Mongolian differs significantly to that of English, with differences in the directionality, moraicity of coda consonants, boundedness, and extrametricality. The complexity of the Mongolian stress system and its dissimilarity to English provides an informative test area for various L2 stress hypotheses. Primary stress placement in Khalkha Mongolian is governed by the following rules: 

        The stress system of Khalkha Mongolian is unbounded, meaning that stress placement is not governed by a binary foot structure. This type of system is referred to as prominence-driven, meaning that stress is assigned based on factors such as syllable weight, edge-alignment, and non-finality (Walker, 1996). This poses even greater difficulty for learners with a footed L1 stress system, requiring that they unlearn and rid their speech of the foot, a process which Özçelik (2016) predicts to be very difficult if not impossible. While constraint-based analyses of the Mongolian stress system have proven to be much more effective in explaining the effects of non-finality and default-opposite-edge prominence mentioned in (17), the approach taken in the present study is a Parametric one, presumably since the L1 of the learners is a bounded system. The rules given in (17) are exemplified below, with H representing a heavy syllable6, and L representing a light syllable (modified from Özçelik, 2021, p. 2): 

        The author found that many of the errors produced by the English speakers were largely defined by their level of experience, or more specifically, their stage of acquisition (Özçelik, 2016). For example, when presented with Mongolian words containing all light syllables, e.g. загасчин [dzá.ɢas.ʧiŋ], novice learners seemingly applied English stress parameters to the word, yielding *[dza.ɢás.ʧiŋ]. The novice learners had presumably not yet learned that coda consonants in Mongolian do not carry morae, which led them to place primary stress on the penultimate closed syllable7. Different learners who made the same errors were later organized into one of five groups, with each group reflecting a stage in the path of acquisition. At each stage, the most common error was ascribed to the resetting of a specific parameter. Learners who incorrectly placed stress on the penultimate syllable of [dza.ɢás.ʧiŋ] were placed into group 1, since none of their parameters were assumed to be reset.  

Compared to the novice learners in stage 1, learners in stage 2 commonly misplaced stress on the last syllable, as in *[dza.ɢas.ʧíŋ]. Stage 2 learners were therefore assumed to have reset the ‘Extrametricality’ parameter from ‘Yes’ to ‘No’ which allowed for End-Rule: Right to assign stress on the rightmost foot.  

Each stage’s most common error is adequately explained through the resetting of a new parameter, ending with correct stress placement in light syllable words in stage 5 learners. The author concluded that the evidence supports the claim that adult learners are able to reset parameters in adulthood and that access to UG persists even following the critical period. This evidence also supports Özçelik’s previous hypothesis that learners follow a specific “path” when acquiring prosodic features of an L2’s phonology, also known as the Prosodic Acquisition Path Hypothesis (Özçelik, 2016), mentioned above. This paper describes one the first investigations in Mongolian stress acquisition, providing valuable insights in learnability, as well as the interaction between foot-based and prominence-based systems in the IL.  


Optimality Theoretic Approaches 

        In one of the earliest OT studies on L2 stress acquisition, Ou (2003) investigates the stress patterns of Mandarin learners of English. Ou claims that Mandarin Chinese is a QI language, which uses tone or pitch to determine lexical contrasts. Contrary to other theories regarding stress rules in Mandarin, the author assumes that Mandarin is a language governed by trochaic feet, as in (má.yi) 蚂蚁. The constraint rankings for Mandarin are presented in (21). 

Mandarin Chinese is a QI language, with no sensitivity to weight in the nucleus or the coda. As seen in the tableau above, WSP is dominated by TROCHAIC, which states that a trochaic foot such as (pí.piŋ) is more harmonic than a candidate which prioritizes stress on a heavy syllable, as in (pi.píŋ). The constraint TROCHAIC is undominated in both English and Mandarin, thus it will not be involved in the discussion regarding constraint reranking. Undominated PARSE-σ eliminates any candidates which are not syllabic trochees, meaning that a foot made up of two syllables, such as (pi.piŋ), is more harmonic than a foot containing only one syllable, as in pi(piŋ). NON-FINAL must also be ranked below PARSE-σ, to allow for footed final syllables. This means that a fully parsed word, (σ σ) (σ σ), is preferred over a word which leaves the final syllable unparsed, as in (σ σ) σ σ, or σ (σ σ) σ.  

        These constraints’ rankings are practically reversed in English, as demonstrated in (22). English stress placement is determined by various other constraints, such as Ft-BINARITY and MAIN-R, however only the relevant constraints relative to Mandarin are presented here. Furthermore, as mentioned above, English extrametricality varies by word class. Nouns are thought to have syllable extrametricality, while verbs typically have consonant extrametricality. Ou frames this difference in extrametricality through two non-finality constraints which differ based on word class. NON-FINAL becomes NON-FINAL(σ) in nouns, and NON-FINAL(C) in verbs. Tableaux for both an English noun and verb are provided in (22). 

Since only a few constraints are involved in the reranking process, namely WSP, PARSE-σ, and NON-FINAL, the possible intermediate stages of an IL can be easily derived. Ou outlines the three hypothetical constraint rankings that learners will have on their way to the correct L2 ranking, thus predicting possible errors. He states that the ranking in (23c) should not be possible before (23a) or (23b), given his assumption that constraint reranking occurs one constraint at a time. The author also assumes that constraint reranking can include the promotion of some constraints, although he does not specifically describe any algorithmic reranking procedures. 

         In order to test whether the hypothesized intermediate stages would realize an output determined by any of the rankings in (23), he presented Mandarin learners of English with the task of deciding which nonce words (nouns vs. verbs) sounded more grammatical8. The two main factors investigated were sensitivity to extrametricality in nouns vs. verbs (determined by the ranking of the non-finality constraint), and sensitivity to syllable weight (determined by the ranking of WSP). When comparing the two factors within speakers, the author found learners who displayed sensitivity to both syllable weight and extrametricality (successful acquisition, 23c), learners who displayed sensitivity to only extrametricality and not to syllable weight (an intermediate stage, 23b), and one learner who displayed sensitivity to neither (the initial L1 state). The results are displayed in (24). 

Learners in an intermediate stage who display sensitivity to syllable weight and not extrametricality were not attested, going against the predicted ranking in (23a). This implies that constraint reranking is restricted, or that certain constraint must be reranked before another is demoted or promoted. Ou claims that this implication warrants further revisions to the existing model of OT constraint reranking, since previous assumptions postulated that all possible rerankings could potentially account for attested variation in errors. A later study conducted by the authors (Ou & Ota, 2015, mentioned above) ascribes the acquisition of extrametricality in nouns and verbs to generalizations or analogy, which could explain the asymmetrical acquisition of extrametricality displayed in this study. Further research is required to better understand the processes which trigger constraint reranking, and whether the acquisition of quantity sensitivity is treated differently than that of extrametricality among other populations of learners. 

        In a recent study investigating the L2 English of Moroccan Arabic (MA) speakers, Mohamed (2021) uses OT to explain learner stress placement errors using the Constraint Demotion Algorithm (Tesar & Smolensky, 2000). After delineating the constraint rankings for either language, he demonstrates that many of the errors in stress placement are attributed to initial transfer of L1 constraint rankings in the IL. The constraint rankings presented for MA (25) and English (26) are shown below: 

For example, in MA the IAMBIC constraint dominates the TROCHAIC constraint, yielding ultimate stress in disyllabic nouns, as seen below, taken from Mohamed (2021, p. 13): 

In disyllabic English nouns (coffee, attic, object), these learners appear to apply their L1 constraint rankings, yielding the suboptimal form in (28a). In order to place stress on the correct syllable, learners would need to demote IAMBIC below TROCHAIC, as seen in (28b). A demoted constraint is represented in boldface in a shaded box. 

        During the learning process, however, as learners begin to rerank their constraints hierarchies one at a time, stress placement patterns emerge which are not attributed to the rankings of either the L1 or the L2. In MA learners, a common stress placement error was the pronunciation of cigarette with penultimate stress, as in cigárette, an error made by 67% of learners (Mohamed, 2021, p. 12). This error cannot be attributed to the L1 constraint rankings of MA presented in (25), as demonstrated below in (29). The unchanged MA constraint rankings would yield the output cigarétte. 

This implies that most Moroccan learners have already begun to alter some of their constraint rankings. As Mohamed argues, the first two constraints demoted to a lower stratum are WSP and IAMBIC, due to their problematic effect on disyllabic English nouns. IAMBIC must be demoted below TROCHAIC to allow for words like ‘coffee’ to carry stress on the penult (cóffee). WSP must be demoted below NON-FINALITY, so that heavy final syllables in words like ‘cigarette’ do not carry primary stress (cigarétte). These constraints are hypothetically “demoted” below all other relevant constraints (not just the constraints they compete with) significantly reducing their impact on the output evaluation. If IAMBIC and WSP are demoted below the remaining constraints shown in (30), then the remaining constraint ranking would yield the attested output from the MA learners. 

        Mohamed continues his analysis by explaining each stress placement error through the demotion of particular constraints, effectively providing evidence that learners exhibit stage-like development in stress acquisition. He goes on to describe the necessary constraint demotions for MA learners to fully acquire the English stress system, therefore creating an outline that can account for learner errors in any stage of acquisition. This study also provides valuable L2 stress acquisition data from speakers of Moroccan Arabic, a relatively underrepresented demographic in the literature. 



This paper presents a survey of selected studies on L2 stress acquisition from various theoretical approaches. This review also provides a brief overview of stress, metrical phonology, Parameter and Optimality Theory, as well as second language phonology in general. In summary, practically every study covered here has expanded on the idea that L2 phonology is largely influenced by the L1, an idea known as transfer. Other studies citing IL phonological patterns that are not attributed to L1 nor the L2 are explained using the ideas of universal constraints on natural phonologies, which includes IL phonologies. Studies using metrical approaches to acquisition have provided insight into whether certain properties of metrical systems are universal, and other studies which do not take a metrical approach provide alternate interpretations of empirical data which contribute to the larger conversation surrounding the role of metrical phonology in L2 stress acquisition. Since the majority of studies conducted on L2 stress acquisition (and L2 phonology in general) are centered on the acquisition of English, further studies in L2 stress would benefit from a more diverse range of language pairings in the L1 and L2. Another subfield of L2 phonology that should be expanded upon is the acquisition of vowel reduction processes, a phenomenon directly related to stress which can be investigated in conjunction with L2 stress acquisition. Lastly, the results presented in this review can potentially have important implications for pedagogical implementation in the classroom and can change the way language instructors teach stress patterns. 




Archibald, J. (1993). Language learnability and L2 phonology: The acquisition of metrical parameters. Dordrecht, Netherlands: Kluwer Academic Publishers.  
Archibald, J. (1994). A formal model of learning L2 prosodic phonology. Second Language Research, 10(2), 215–240. 

Archibald, J. (1998). Metrical parameters and lexical dependency: Acquiring L2 stress. In S. Flynn, G. Martohardjono, & W. O’Neil (Eds.), The    Generative Study of Language Acquisition (pp. 279–302). Mahwah, NJ: Lawrence Erlbaum Assoc. 

Bassetti, B. (2006). Orthographic Input and Phonological Representations in Learners of Chinese as a Foreign Language. Written Language and Literacy. 9(1), 95-114. 

Best, C. (1993). Emergence of language-specific constraints in perception of non-native speech: A window on early phonological development. In B. de Boysson-Bardies (Ed.), Developmental neurocognition: Speech and face processing in the first year of life (pp. 289–304). Dordrecht: Kluwer. 

Best, C. (1993). Emergence of language-specific constraints in perception of non-native speech: A window on early phonological development. In B. de Boysson-Bardies (Ed.), Developmental neurocognition: Speech and face processing in the first year of life (pp. 289–304). Dordrecht: Kluwer. 

Best, C. (1994). The emergence of native-language phonological influence in infants: A perceptual assimilation model. In H. Nussbaum, J. Goodman, & C. Howard (Eds.), The transition from speech to spoken words: The development of speech perception (pp. 167–224). Cambridge, MA: MIT Press. 

Best, C. (1995). A direct realist view of cross-language speech perception. In W. Strange (Ed.), Speech perception and linguistic experience: Issues in cross-language research (pp. 171–204). Timonium, MD: York Press. 

Boersma, Paul (1997). How we learn variation, optionality, and probability. IFA Proceedings. 21:43-58. 

Broselow, E., Chen, S., & Wang, C. (1998). The emergence of the unmarked. Studies in Second Language Acquisition, 20(2): 261–280. 

Brulard, I. & Carr, P. (2003). French–English bilingual acquisition of phonology: One production system or two? International Journal of Bilingualism, 7(2), 177–202. 

Carlson, L. (1978). Word Stress in Finnish. Unpublished manuscript, MIT. 

Chomsky, N. (1981a). Lectures on Government and Binding. Dordrecht: Foris. 

Chomsky, N. (1981b). Principles and parameters in syntactic theory. In N. Hornstein & D. Lightfoot (Eds.), Explanation in linguistics (pp. 32–75). London: Longman. 

De Houwer, A. (2009). Bilingual First Language Acquisition. Tonawanda, NY: Multilingual Matters. 

Dell, F. (1984). L’accentuation dans les phrases en français. In F. Dell, D. Hirst & J. R. Vergnaud (Eds.), Forme sonore du langage (pp. 65-122). Paris: Hermann. 

Dresher, B. E., & Kaye, J. (1990). A computational learning model for metrical phonology. Cognition, 34, 137–195. 

Dupoux, E., Peperkamp, S. & Sebastián-Gallés, N. (2001) A robust paradigm to study stress ‘deafness’. Journal of the Acoustical Society of America, 110, 1606-1618. 

Eckman, F. R., & Iverson, G. K. (1994). Pronunciation difficulties in ESL: Coda consonants in English interlanguage. In M. Yavaş (Ed.), First and second language phonology (pp. 251–265). San Diego, CA: Singular. 

Eckman, F. R. (1977). Markedness and the contrastive analysis hypothesis. Language Learning, 27, 315–330. 

Eckman, F. R. (2008). Typological markedness and second language phonology. In J. G. Hansen Edwards & M. L. Zampini (Eds.), Phonology and Second Language Acquisition (pp. 95–116). Philadelphia, PA: John Benjamins. 

Flege, J. E. (1987). The production of new and similar phones in a foreign language: Evidence for the effect of equivalence classification. Journal of Phonetics, 15, 47–65. 

Flege, J. E. (1995). Second language speech learning: Theory, findings, and problems. In W. Strange (Ed.), Speech perception and linguistic experience: Issues in cross-language research (pp. 233–277). Timonium, MD: York Press. 

Garcia, G. D., & Guzzo, N. B. (2017). The acquisition of English stress by Québec francophones. In M. Yavaş, M. Kehoe, & W. Cardoso (Eds.), Romance-Germanic Bilingual Phonology (pp. 200–221). Sheffield: Equinox. 

Goad, H., & Prévost, A. E. (2011). A test case for markedness: The acquisition of Québec French stress. Unpublished manuscript. McGill University. 

Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds "l" and "r." Neuropsychologia, 9, 317–323. 

Halle, M., & Vergnaud, J. R. (1987). An Essay on Stress. MIT Press. 

Hammond, M. (1995). Metrical Phonology. Annual Review of Anthropology, 24, 313–342. 

Hancin-Bhatt, B. (1994). Segment transfer: A consequence of a dynamic system. Second Language Research, 10, 241–269. 

Hancin-Bhatt, B. (2000). Optimality in second language phonology: Codas in Thai ESL. Second Language Research, 16(3): 201–232. 

Hancin-Bhatt, B., & Bhatt, R. (1997). Optimal L2 syllables: Interaction of transfer and developmental effects. Studies in Second Language Acquisition, 19: 331–378. 

Hao, Y. C. (2012). Second language acquisition of Mandarin Chinese tones by tonal and non-tonal language speakers. Journal of Phonetics. 40(2), 269-279. 

Harris, J. (1983). Syllable structure and stress in Spanish. Cambridge, MA: MIT Press. 

Hayes, B. (1981). A metrical theory of stress rules. Doctoral dissertation, MIT. 

Hayes, B. (1982). Extrametricality and English stress. Linguistic Inquiry, 13, 227–276. 

Hayes, B. (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry, 20, 253–306. 

Hayes, B. (1995). Metrical stress theory: Principles and case studies. Chicago & London: The University of Chicago Press. 

Hayes, B., & Puppel, S. (1985). On the rhythm rule in Polish. In H. van der Hulst & N. Smith (Eds.), Advances in Nonlinear Phonology (pp. 59–81). Dordrecht: Foris. 

Hayes-Harb, R., & Hacking, J. F. (2015). The influence of written stress marks on native English speakers' acquisition of Russian lexical stress contrasts. Slavic and East European Journal, 59(1), 91–109. 

Hayes-Harb, R., Nicol, J., & Barker, J. (2010). Learning the Phonological Forms of New Words: Effects of Orthographic and Auditory Input. Language and Speech. 53(3), 367-81. 

Hyman, L. (1984). A theory of phonological weight. Dordrecht: Foris. 

Ioup, G. (2008). Exploring the role of age in the acquisition of a second language phonology. In J. G. Hansen Edwards & M. L. Zampini (Eds.), Phonology and Second Language Acquisition (pp. 41–62). Philadelphia, PA: John Benjamins. 

Kager, R. (1999). Optimality Theory. Cambridge: Cambridge University Press. 

Kelly, M. H., & Bock, J. K. (1988). Stress in time. Journal of Experimental Psychology: Human Perception and Performance, 14, 389–403. 

Kerek, A. (1971). Hungarian Metrics: Some Linguistic Aspects of Iambic Verse. Indiana University Publications, Uralic and Altaic Series, 117. Mouton, The Hague. 

Lado, R. (1957). Linguistics across cultures: Applied linguistics for language teachers. Ann Arbor: University of Michigan Press. 

Liberman, M., & Prince, A. (1977). On stress and Linguistic Rhythm. Linguistic Inquiry, 8, 249–336. 

Lombardi, L. (2003). Second language data and constraints on Manner: Explaining substitutions for the English interdentals. Second Language Research, 19: 225–250. 

Major, R. C. (2001). Foreign accent: The ontogeny and phylogeny of second language phonology. Mahwah, NJ: Lawrence Erlbaum Associates. 

Major, R. C. (2008). Transfer in second language phonology: A review. In J. G. Hansen Edwards & M. L. Zampini (Eds.), Phonology and Second Language Acquisition (pp. 63–94). Philadelphia, PA: John Benjamins. 

McCarthy, J., & Prince, A. (1986). Prosodic morphology. Unpublished manuscript, University of Massachusetts, Amherst & Brandeis University. 

Mohamed, S. (2021). An Optimality-Theoretic Analysis of Word-stress: Evidence from Moroccan-English Interlanguage. Unpublished manuscript. Ibn Tofail University, Kenitra.  

Ou, S., & Ota, M. (2015). Is second-language stress acquisition guided by metrical principles? Evidence from Mandarin-speaking learners of English. In Y. E. Hsiao & L. H. Wee (Eds.), Capturing phonological shades within and across languages (pp. 389–413). Cambridge, UK: Cambridge Scholars Publishing. 

Özçelik, Ö (2016). The Prosodic Acquisition Path Hypothesis: Towards explaining variability in L2 acquisition of phonology. Glossa, 1(1), Article 28. 

Özçelik, Ö. (2018). Universal Grammar and second language phonology: Full Transfer / Prevalent Access in the L2 Acquisition of Turkish "Stress" by English and French Speakers. Language Acquisition, 25(3), 231–267. 

Özçelik, Ö. (2021). L2 Acquisition of a Complex Stress Pattern: UG-Constrained Learning Paths in Khalkha Mongolian. Frontiers in Psychology, 12, 627-797. 

Pater, J. (1997). Metrical parameter missetting in second language acquisition. In S. J. Hannahs and M. Young-Scholten (Eds.), Focus on Phonological Acquisition (pp. 235–261). Amsterdam: John Benjamins. 

Peperkamp, S. & Dupoux, E. (2002). A typological study of stress ‘deafness’. In C. Gussenhoven & N. Warner (Eds.), Laboratory Phonology 7 (pp. 203-240). Berlin, New York: De Gruyter Mouton. 

Prince, A. (1990). Quantitative consequences of rhythmic organization. In K. Deaton, M. Noske, & M. Ziolkowski (Eds.), Proceedings of the Chicago Linguistics Society 26-II (pp. 471–490). Chicago: Chicago Linguistic Society. 

Prince, A., & Smolensky, P. (1993). Optimality Theory: Constraint interaction in generative grammar. Unpublished manuscript. Rutgers University & University of Colorado, Boulder. 

Scovel, T. (1988). A time to speak: A psycholinguistic inquiry into the critical period for human speech. Cambridge, MA: Newbury House. 

Schwartz, B. D., & Sprouse, R. A. (1996). L2 cognitive states and the full transfer/full access model. Second Language Research, 12, 40–72. 

Selinker, L. (1972). Interlanguage. IRAL, 10, 209–231. 

Selkirk, E. (1980). The role of prosodic categories in English word stress. Linguistic Inquiry, 11, 563–605. 

Sheldon, A., & Strange, W. (1982). The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that speech production can precede speech perception. Applied Psycholinguistics, 3, 243–261. 

Smith, N. V. (1973). The acquisition of phonology: A case study. New York: Cambridge University Press. 

Strozer, Judith R. (1994). Language Acquisition after Puberty. Washington, D.C.: Georgetown University Press. 

Tesar, B. (1997). An Iterative Strategy for Learning Metrical Stress in Optimality Theory. In Proceedings of the 21st Annual Boston University Conference on Language Development (pp. 615–626). Cascadilla Press. 

Tesar, B., & Smolensky, P. (2000). Learnability in Optimality Theory. Cambridge, MA: MIT Press. 

Thu T. A. Nguyen, & Ingram, J. (2005). Vietnamese Acquisition of English Word Stress. TESOL Quarterly, 39(2), 309–319. 

van der Pas, B. & Zonneveld, W. (2004). L2 parameter resetting for metrical systems (An assessment and a reinterpretation of some core literature). The Linguistic Review, 21, 125-170. 

Vogel, I. (2009). Universals of prosodic structure. In S. Scalise, E. Magni, & A. Bisetto (Eds.), Universals of language today. (pp. 59–82). Dordrecht: Springer. 

Walker, R. (1996). Prominence-Driven Stress. Unpublished manuscript. University of California, Santa Cruz. 

Walker, R. (1997). Mongolian stress, licensing, and factorial typology. Unpublished manuscript. University of California, Santa Cruz. 

White, L. (2003). Second language acquisition and universal grammar. Cambridge, UK: Cambridge University Press. 

Youssef, A. & Mazurkewich, I. (1998). The acquisition of English metrical parameters and syllable structure by adult native speakers of Egyptian Arabic (Cairene Dialect). In S. Flynn, G. Martohardjono, & W. O’Neil (Eds.), The Generative Study of Language Acquisition (pp. 303–332). Mahwah, N.J.: Lawrence Erlbaum Assoc. 

Zec, D. (1995). Sonority Constraints on Syllable Structure. Phonology, 12(1), 85–129. 

Zhang, H. (2016). Dissimilation in the second language acquisition of Mandarin Chinese tones. Second Language Research, 32(3), 427-451. 

This page has paths:

This page references: