Phonological priming

Research on word recognition and lexical processing often relies on the method of phonological priming to explore how words are selected in the mental lexicon based on their phonological form over other similar-sounding words (Dufour & Peereman, Reference Dufour and Peereman2003a, Reference Dufour and Peereman2003b, Reference Dufour and Peereman2004, Reference Dufour and Peereman2009; Goldinger, Luce & Pisoni, Reference Goldinger, Luce and Pisoni1989; Hamburger & Slowiaczek, Reference Hamburger and Slowiaczek1996; Monsell & Hirsh, Reference Monsell and Hirsh1998; Radeau, Morais & Segui, Reference Radeau, Morais and Segui1995; Slowiaczek & Hamburger, Reference Slowiaczek and Hamburger1992). This method has been used in the study of phonological encoding of nonnative words, and lexical competition in nonnative word recognition (Barrios, Jiang & Idsardi, Reference Barrios, Jiang and Idsardi2016; Broersma & Cutler, Reference Broersma and Cutler2008, Reference Broersma and Cutler2011; Broersma, Reference Broersma2012; Cook, Reference Cook2012; Cook & Gor, Reference Cook and Gor2015; Cook, Pandža, Lancaster & Gor, Reference Cook, Pandža, Lancaster and Gor2016; Darcy, Daidone & Kojima, Reference Darcy, Daidone and Kojima2013; Darcy, Dekydspotter, Sprouse, Glover, Kaden, McGuire & Scott, Reference Darcy, Dekydspotter, Sprouse, Glover, Kaden, McGuire and Scott2012; Gor & Cook, unpublished manuscript; Pallier, Colomé & Sebastián-Gallés, Reference Pallier, Colomé and Sebastián-Gallés2001).

Phonological priming is a version of form priming, in which two phonologically overlapping words are presented, and the word presented first (the prime) influences the accuracy and speed of responses to the second word (the target). For example, upon hearing the prime GRIEF, listeners take longer to respond to the target GREEN compared to the unrelated condition with the prime CLUMP (Slowiaczek & Hamburger, Reference Slowiaczek and Hamburger1992). Note that faster processing is referred to as facilitation, and slower processing as inhibition. In auditory priming both the prime and the target are auditory, while in cross-modal priming the prime is auditory and the target is visual. The advantage of auditory priming is that it relies on phonological representations, and does not involve any graphic-phonological conversion necessitated by the use of orthography. At the same time, cross-modal priming has the advantage of being modality-free, and is believed to access central representations of lexical items in the mental lexicon (Dufour, Reference Dufour2008). The direction of the phonological priming effect – facilitation or inhibition – depends on a number of factors (see Dufour, Reference Dufour2008 for a review). In general, phonological inhibition is interpreted as a manifestation of lexical competition between phonologically overlapping words, while phonological facilitation is interpreted as a manifestation of a processing boost provided by the overlap in smaller phonological segments (phonemes or syllables) at the sublexical level (Slowiaczek & Hamburger, Reference Slowiaczek and Hamburger1992). This review will address the factors relevant for the interpretation of the phonological priming experiments with nonnative speakers discussed below.

Phonological priming effects have been studied with the help of two versions of a lexical decision task (LDT) with phonological priming (further referred to as a phonological priming task): LDT with short-lag and medium-lag priming. In short-lag phonological priming experiments, participants hear the prime quickly followed by the target, and are asked to perform a lexical decision on the target: decide whether the target is a real word or a nonword, and press one of the two response keys corresponding to these two choices as quickly and accurately as possible. Accordingly, half of the targets in LDT are nonwords added to real words in order to make the task meaningful, and more filler trials are added to make the purpose of the task less obvious to participants. The interval between the prime and the target, or the interstimulus interval (ISI) varies across experiments, and is often round 250–350 milliseconds (ms).Footnote ¹ In medium-lag priming, the prime and the target are separated by several intervening words, for example, 8–20 (Pallier et al., Reference Pallier, Colomé and Sebastián-Gallés2001). In the medium-lag version of the task, stimuli are presented continuously, and not in pairs, as in short-lag priming, and responses are given to each stimulus, which happen to be primes, targets, or fillers.

The phonological priming task has been used to explore three issues at the interface of nonnative phonology and lexical processing. First, it has been used in research on nonnative phonological perception to establish how problems with phonological categorization of L2 sounds impact the phonological encoding and auditory recognition of L2 words. Second, it has been used to study the impact of problems with phonological categorization of L2 sounds on lexical competition in L2 word recognition. And third, it has been used to investigate a new issue in L2 lexical processing that has recently attracted researchers’ attention: unfaithful, or fuzzy representations of low-familiarity L2 words. This direction explores the idea that because L2 speakers receive reduced input, low-familiarity words do not have robust representations, even when no difficult L2 phonological contrasts are involved. We will review the existing phonological priming research addressing these three issues.

Problems with phonological categorization lead to problems with phonological encoding of L2 words

Nonnative speakers, who often have low sensitivity to phonological contrasts in L2 that are absent in their native language, also experience problems with L2 word recognition when these difficult contrasts are involved (Chrabaszcz & Gor, Reference Chrabaszcz and Gor2017; Cutler, Weber & Otake, Reference Cutler, Weber and Otake2006; Hayes-Harb & Masuda, Reference Hayes-Harb and Masuda2008; Weber & Cutler, Reference Weber and Cutler2004). The direction of acquisition – whether L2 phonological categories are acquired first, and the lexicon is acquired later, or conversely, the lexicon is acquired first – is still a matter of debate. Given the existing research, it is safe to assume that bottom-up (‘categories first’) and top-down (‘lexicon first’) acquisition interact in L2 in a bootstrapping fashion; however, the establishment of L2 phonological categories is a prerequisite for the development of robust phonological representations of numerous L2 words stored in the mental lexicon (see Gor, Reference Gor and Schwieter2015 for a review). The research targeting phonological representations of words, or phonolexical representations (Cook et al., Reference Cook, Pandža, Lancaster and Gor2016), focuses primarily on L2 words with difficult L2 contrasts. It makes use of phonological priming combined with a LDT, and compares the effects of medium-lag repetition priming (facilitation on the second presentation of the same real word), and the priming in minimal pairs of words contrasted by one phoneme. The medium-lag repetition priming experiments, which test nonnative phonological sensitivity, use a continuous presentation mode, with real-word primes and targets separated by several intervening words or nonwords (Barrios et al., Reference Barrios, Jiang and Idsardi2016; Darcy et al., Reference Darcy, Dekydspotter, Sprouse, Glover, Kaden, McGuire and Scott2012; Pallier et al., Reference Pallier, Colomé and Sebastián-Gallés2001). Both native and nonnative listeners respond faster on the second presentation of the same word. However, differences emerge for minimal pairs of words differentiated by a difficult L2 phonemic contrast – nonnative speakers show facilitation for such prime-target pairs as well. This facilitation is interpreted as an indicator of nonnative insensitivity to the phonological contrasts that differentiate these minimal word pairs. If two words that constitute minimal pairs and differ by a difficult phonemic contrast, the prime and the target, and show facilitation in phonological priming, this signals that they are treated as a repetition of the same word by nonnative listeners. Accordingly, such facilitation in L2 is a sign of lexical confusion resulting from phonological insensitivity, which leads to unfaithful phonolexical representations. Studies of L2 phonological perception use medium-lag repetition priming to test the faithfulness of phonolexical encoding of the prime, and predict facilitation for minimal pairs in nonnative speakers, in the absence of facilitation in native speakers. The first study to use medium-lag repetition priming compared Catalan identical pairs of words and nonwords, and minimal pairs differing by a Catalan-specific contrast – /e/-/ɛ/, /o/-/ɔ/, and /s/-/z/ – absent in Spanish. Spanish-dominant, but not Catalan-dominant bilinguals showed facilitation not only for identical pairs, but also for minimal pairs of words, which were separated by 8–20 intervening words in a continuous presentation. The authors interpreted this finding as an indication that minimal pairs were treated as homophones because of insensitivity of Spanish-dominant bilinguals to the Catalan-specific phonological contrasts.

Darcy and colleagues (Reference Darcy, Dekydspotter, Sprouse, Glover, Kaden, McGuire and Scott2012) used the same medium-lag repetition priming method to compare the processing of French minimal pairs of words differing by the /y/-/u/, /œ/-/ɔ/ and /y/-/i/ vowels in native speakers and intermediate and advanced English learners of French. They found that intermediate French learners showed spurious facilitation for minimal pairs of words contrasted by the /y/ and /u/ vowels, but not the other vowels. Advanced learners did not show problems with the lexical encoding of minimal pairs differentiated by any of the three contrasts absent in English. Remarkably, the results in two nonnative groups did not align with the results of a phonetic categorization task, which suggests that it may sometimes be possible to acquire lexical contrasts without the underlying phonological contrasts. Most likely, nonnative speakers acquire the sound form of high-frequency L2 words, and may store their gestalt sound representations even before all the phonemes and contrasts are firmly established.

Finally, a medium-lag repetition priming experiment reported that advanced Spanish learners of English did not acquire phonolexical representations of English minimal pairs of words differentiated by the vowel contrasts /ɑ/–/æ/ and /i/–/ɪ/ absent in Spanish. Advanced Spanish learners of English, but not English speakers showed facilitation for the minimal pairs as well as the same words repeated with 9–23 intervening words (Barrios et al., Reference Barrios, Jiang and Idsardi2016). Importantly, an asymmetry in nonnative facilitation was observed with regard to the sequence of words constituting minimal pairs with the vowels /ɑ/–/æ/ and /i/–/ɪ/ in the experimental lists. Minimal pairs resulted in facilitation in Spanish learners of English only in that sequence, i.e., when the word beat /bit/ preceded bit /bɪt/, but not when bit preceded beat. Such an asymmetry had been previously observed in a visual world eye-tracking study targeting lexical competition in Dutch L2 learners of English with another vowel contrast, /æ/–/ɛ/, as in panda and pencil (Weber & Cutler, Reference Weber and Cutler2004). When the target was panda, participants also looked at the picture of the competitor pencil, but when the target was pencil, they did not look at the picture of panda. Asymmetrical lexical access is an argument against complete homophony of such minimal pairs in nonnative processing – a claim made earlier by Pallier and colleagues (Reference Pallier, Colomé and Sebastián-Gallés2001). Furthermore, this asymmetry indicates that words with an ‘easier’ vowel that has a close counterpart in L1 are better encoded and more efficiently accessed in the nonnative mental lexicon, and as a result, prime their minimal pairs stronger. Overall, medium-lag priming studies show that problems with phonological categorization percolate to phonolexical encoding of L2 words, and lead to phonolexical ambiguity (see Chrabaszcz & Gor, Reference Chrabaszcz and Gor2017).

Weak lexical competition of phonolexically ambiguous nonnative words

The second direction of phonological priming research explores how L2 speakers deal with phonolexical ambiguity, which results from problems with phonological categorization, at the level of lexical competition. It does so by establishing the direction of the priming effects when the primes and targets with initial overlap are presented consecutively, with a short interstimulus interval. In short-lag phonological priming with native listeners, one-phoneme initial overlap, as in GOALS-GREEN, leads to facilitation (Hamburger & Slowiaczek, Reference Hamburger and Slowiaczek1996; Slowiaczek & Hamburger, Reference Slowiaczek and Hamburger1992). Such facilitation is explained by a sublexical processing boost that the prime gives to the target, before lexical representations of words are activated and start competing for selection. By contrast, word-initial three-phoneme overlap leads to inhibition (lower accuracy and longer reaction times (RTs) in native speakers of English (Hamburger & Slowiaczek, Reference Hamburger and Slowiaczek1996; Slowiaczek & Hamburger, Reference Slowiaczek and Hamburger1992) and French (Dufour & Peereman, Reference Dufour and Peereman2003b). Note that both initial overlap and final mismatch are needed to create inhibition (Dufour & Peereman, Reference Dufour and Peereman2004). Word-initial phonological inhibition is associated with lexical competition, primarily, between the prime and the target, but also with phonological neighbors, or similar-sounding words activated by the prime and the target during lexical selection.Footnote ²

Several cross-modal priming studies with auditory primes and visual targets explore what they refer to as phantom activation, or spurious activation of irrelevant phonological neighbors in nonnative lexical access (Broersma & Cutler, Reference Broersma and Cutler2008, Reference Broersma and Cutler2011; Broersma, Reference Broersma2012). According to Broersma and Cutler, nonnative listeners, who do not have a robust representation of a phonological contrast, incorrectly activate irrelevant words differentiated by this contrast. For example, Dutch listeners have problems with contrasting English phonemes /æ/ and /ɛ/ (Weber & Cutler, Reference Weber and Cutler2004). The studies by Broersma and Cutler target vowel contrasts in word onsets in such a way that for English speakers, the vowel phonemes in two words do not match, while for Dutch speakers, they may incorrectly match, which will create an initial overlap in several phonemes. English words DEFinite and DAFfodil were recorded, and their onsets corresponding to the word ‘deaf’ and near-word ‘daf’ excised and presented as auditory primes. The words DEFinite and DAFfodil served as visual targets. While native English speakers showed facilitation only for the matching pairs, e.g., daf-DAFfodil, nonnative speakers also showed facilitation across the pairs, e.g., daf-DEFinite (Broersma & Cutler, Reference Broersma and Cutler2011).Footnote ³ Another study replicated these findings and interpreted them as activation of mismatching lexical representations that occurred more often in L2 than L1 listeners (Broersma, Reference Broersma2012). It also tested the L1-L2 differences in lexical competition. A second cross-modal priming experiment showed the expected facilitation for identical prime-target pairs flesh-FLESH in both L1 and L2, but consistent inhibition for mismatching flesh-FLASH only in L1 participants. L2 listeners in most cases showed facilitation for mismatching words. This is another set of findings documenting facilitation resulting from phonological confusion and problems with phonolexical encoding in L2 (also discussed in the framework of the lexical coding deficiency hypothesis, see Darcy et al., Reference Darcy, Daidone and Kojima2013).

If indeed phantom activation increases the number of activated competitors, then the strength of competition for lexical selection should also increase, and this should lead to inhibition in priming experiments. The reported effects for L2 are the opposite: facilitation (Broersma, Reference Broersma2012). The research belonging to the third direction discussed below suggests a possible explanation for this conflict by evoking the notion of fuzzy phonolexical representations (Cook, Reference Cook2012; Cook & Gor, Reference Cook and Gor2015; Cook et al., Reference Cook, Pandža, Lancaster and Gor2016; Darcy et al., Reference Darcy, Daidone and Kojima2013). The initial premise is that phonolexical representations of L2 words with perceptually difficult phonemes are not properly encoded in the mental lexicon, and are phonolexically underdifferentiated, or fuzzy. Words with fuzzy representations are weak competitors, and therefore, even if their number is increased due to phantom activation, the activation of individual words with fuzzy representations is low, and they are not strongly engaged in lexical competition. In contrast, activation of sublexical units takes over, and it leads to facilitation, which may increase for phonotactically highly probable phoneme clusters (Vitevitch & Luce, Reference Vitevitch and Luce1999, Reference Vitevitch and Luce2016).

Less familiar nonnative words have weak, underdifferentiated, or fuzzy phonolexical representations

The third direction of research that makes use of phonological priming explores the claim that fuzzy phonolexical representations is a more general property of less familiar nonnative words, and such words become weak competitors because they have weak or fuzzy phonolexical representations even if they do not contain phonologically difficult segments. This hypothesis regarding auditory representations of nonnative words in the mental lexicon is based on the Lexical Quality hypothesis proposed for orthographic representations in reading (Perfetti, 2007). According to Perfetti, words are differently entrenched in the mental lexicon depending on their frequency in the input. To extend the Lexical Quality hypothesis to auditory speech, given the reduced nonnative input and inefficient encoding of L2 sounds, low-frequency less familiar words are weakly entrenched and have fuzzy representations, phonolexical as well as orthographic. To test this claim, nonnative priming effects were examined in short-lag auditory priming experiments with a three-phoneme initial overlap between the prime and the target (Cook, Reference Cook2012; Cook & Gor, Reference Cook and Gor2015; Gor, forthcoming; Gor & Cook, unpublished manuscript; Gor, Cook & Jackson, Reference Gor, Cook and Jackson2010). In native speakers, these conditions lead to lexical competition between the prime and the target, and also between phonological neighbors and the target, which results in an inhibition effect. However, a lack of inhibition, and even facilitation was reported for advanced L2 learners of Russian with L1 English. Importantly, nonnative facilitation was observed for less familiar low-frequency primes, which is consistent with weaker lexical competitors having underdifferentiated, or fuzzy phonolexical representations (Cook & Gor, Reference Cook and Gor2015; Gor & Cook, unpublished manuscript; Gor, Cook & Jackson, Reference Gor, Cook and Jackson2010).Footnote ⁴ In particular, Cook and Gor (Reference Cook and Gor2015) report overall inhibition in a LDT with phonological priming in advanced L2 learners of Russian. However, when the reaction time data were reanalyzed separately for unknown, well-known and recognizable primes based on the results of a separate translation task, the trials with well-known primes continued to show inhibition, while the trials with recognizable primes showed a tendency to facilitation. Therefore, the level of familiarity with the prime, presumably related to the level of its entrenchment (Perfetti, 2007), results in either strong competition (inhibition for well-known primes) or weak competition (absence of inhibition for recognizable primes). A similar approach to testing the effect of familiarity of nonnative words, both primes and targets, in a phonological priming experiment relied on corpus-based lexical (lemma) frequency as an estimate of word familiarity (Gor & Cook, unpublished manuscript; Gor, Cook & Jackson, Reference Gor, Cook and Jackson2010). L2 speakers of Russian showed inhibition for high-frequency primes and targets, which were presumably highly familiar to the participants, but facilitation for low-frequency primes and targets, which were less familiar. Importantly, the same nonnative participants showed significant facilitation for the primes and targets from the same high- and low-frequency ranges in a separate semantic priming task, which signals that they were aware of their meaning, i.e., they were familiar with words in these frequency ranges. Therefore, there is emerging evidence supporting the idea that low-familiarity L2 words are weak competitors because they have weak or fuzzy phonolexical representations.

Finally, research on nonnative phonological priming needs to take into account several factors specific for nonnative processing: the level of L2 proficiency of the participants, nonnative lexical frequencies, and the possible role of L1 phonotactics in processing L2 words. While research on phonological priming shows that L2 proficiency plays a role in the development of robust phonolexical representations (Darcy et al., Reference Darcy, Daidone and Kojima2013), even early Spanish–Catalan bilinguals continue to experience problems with phonolexical encoding of minimal pairs of words differentiated by Catalan-specific contrasts (Pallier et al., Reference Pallier, Colomé and Sebastián-Gallés2001). Therefore, the developmental trajectory for individual languages in contact and phonological contrasts may vary, which calls for further research.Footnote ⁵ Also, nonnative lexical frequencies differ from native frequencies because L2 speakers have smaller mental lexicons and lower levels of input and exposure to L2 words (Diependaele, Lemhöfer & Brysbaert, Reference Diependaele, Lemhöfer and Brysbaert2013). Nonnative corpora are currently developed for several languages, and they will allow researchers to make more accurate estimates of nonnative lexical frequencies in the future. Finally, there are indications in the literature that not only L1 phoneme inventories and phonological contrasts influence L2 lexical processing, but phonotactic constraints may also be activated across languages in bilinguals (Freeman, Blumenfeld & Marian, Reference Freeman, Blumenfeld and Marian2016).

To summarize, several future research venues emerge from the accumulated findings regarding native and nonnative phonological priming:

• • In medium-lag phonological priming, identity priming produces facilitation in both native and nonnative speakers, while minimal pairs of words differentiated by a difficult phonological contrast produce facilitation only in nonnative speakers. This nonnative facilitation makes it possible to study the robustness of phonolexical encoding of difficult phonological contrasts in L2, and phonolexical representations of nonnative words.

• • In short-lag phonological priming, word-initial phoneme overlap sufficient to activate lexical competitors (e.g., three-phoneme overlap in English four- or five-phoneme primes and targets) produces inhibition in native speakers, but facilitation for less familiar low-frequency words in nonnative speakers. Since phonological inhibition results from lexical competition, while facilitation results from sublexical processing with phonotactically based expectations, these two effects can be used to identify the type of processing in nonnative speakers: lexical or sublexical.

• • Given that phonological facilitation has a sublexical locus, while inhibition has a lexical locus, facilitation as a sign of weak lexical competition points to weak phonolexical representations of nonnative words, regardless of their phonological difficulty. Consequently, the direction of the phonological priming effect with initial overlap between the prime and the target in at least three phonemes can be used as a diagnostic tool to test the robustness of phonolexical representations of nonnative words.

These observations invite further research on phonological priming in the future study of phonological and lexical processing and their interface in nonnative speakers.