Phonological neighborhood competition affects spoken word production irrespective of sentential contextby Neal P. Fox, Megan Reilly, Sheila E. Blumstein

Journal of Memory and Language


Multilayered and columnar competitive networks for spoken word recognition

S. Kurogi, T. Sasaki, M. Itakura, T. Nishida

The Dynamics of Lexical Competition During Spoken Word Recognition

James S. Magnuson, James A. Dixon, Michael K. Tanenhaus, Richard N. Aslin

Morphology by itself in planning the production of spoken words

Ardi Roelofs, Harald Baayen


tio ten tein ces, Bo

I 0291 revision received 1 April 2015


Speech production lexically-conditioned phonetic variation arises. In Experiment 1, participants produced words

Goldrick (2009) showed that words that have word-initial voicing minimal pairs (MP words; e.g., cape, because gape is an English word) were produced with significantly arborough, 2010, obust, mor that gove influence of lexical-level information on acoustic?ph encoding during spoken word production. Put a way, it is clear that phonologically similar neighbor ence the production of a target word, but less well understood are the conditions that determine whether and/or when neighbors can influence different aspects of spoken word production. A better understanding of this issue 0749-596X/ 2015 Elsevier Inc. All rights reserved. ? Corresponding author. Fax: +1 (401) 863 2255.

E-mail addresses: (N.P. Fox), megan_reilly@ (M. Reilly), (S.E. Blumstein).

Journal of Memory and Language 83 (2015) 97?117

Contents lists available at ScienceDirect

Journal of Memory journal homepage: www.ewhich a word?s acoustic properties are influenced by lexical-level information such as its frequency or its relation to other words, a phenomenon known as lexically-conditioned phonetic variation. For instance, Baese-Berk and 2007; Munson & Solomon, 2004; Sc 2012, 2013; Wright, 2004).

While this general class of findings is r is needed to delineate the parameterse work rn the onetic nother s influ-The phonetic realization of a segment varies considerably depending on many factors, including properties of the word in which the segment occurs. Investigating the sources of such systematic variation can reveal features of the cognitive architecture underlying spoken language production. Recent research has explored the ways in pared to words that do not have word-initial voicing minimal pairs (NMP words; e.g., cake, ?/geIk/ is not an English word). This result exemplifies a body of work indicating that how a target word is pronounced depends on its phonological relationships to other words in the lexicon; the mere existence of similar words in a language can influence acoustic dimensions of a spoken target (Munson,Voice-onset time (VOT)

Phonological competition

Neighborhood density

Sentence context


Introductionin isolation. Results showed that the voice-onset time (VOT) of a target?s initial voiceless stop was predicted by its overall neighborhood density, but not by its having a voicing minimal pair. In Experiment 2, participants read aloud the same targets after semantically predictive sentence contexts and after neutral sentence contexts. Results showed that, although VOTs were shorter in words produced after predictive contexts, the neighborhood density effect on VOT production persisted irrespective of context. These findings suggest that global competition from a word?s neighborhood affects spoken word production independently of contextual modulation and support models in which activation cascades automatically and obligatorily among all of a selected target word?s phonological neighbors during acoustic?phonetic encoding.  2015 Elsevier Inc. All rights reserved. longer voice-onset times (VOTs) for their initial stops com-Article history:

Received 11 July 2014

Two experiments examined the influence of phonologically similar neighbors on articulation of words? initial stop consonants in order to investigate the conditions under whichPhonological neighborhood competi word production irrespective of sen

Neal P. Fox a,?, Megan Reilly a, Sheila E. Blums aBrown University, Department of Cognitive, Linguistic & Psychological Scien bBrown University, Brown Institute for Brain Science, Box 1953, Providence, R a r t i c l e i n f o a b s t r a c tn affects spoken tial context a,b x 1821, Providence, RI 02912, USA 2, USA and Language l sevier .com/locate / jml 98 N.P. Fox et al. / Journal of Memory and Language 83 (2015) 97?117would provide important constraints for models of spoken word production. The current study aimed to examine this issue by probing two questions regarding the conditions under which phonological neighbors can influence a target word?s initial voicing.

Firstly, to what extent is a neighbor?s influence on a target?s production contrast-specific? In particular, can the degree of lengthening of the VOT of the initial /k/ in cape (cf. Baese-Berk & Goldrick, 2009) be attributed to the existence of a voicing contrast in the lexicon (gape), or, alternatively, to the combined influence of its entire neighborhood (gape, tape, cane, cope. . .)? If VOT is associated with the presence or absence of a minimal pair competitor that contrasts on initial voicing, this could lend support to models of spoken word production in which articulatory implementation of a target word is highly dependent on its particular phonological relationships to each of its neighbors. On the other hand, the latter case would support models in which all phonologically related neighbors (rather than just a particularly proximal competitor) contribute to the processes that drive variation in the articulation of voicing.

Secondly, to what extent is a neighbor?s influence on a target?s production context-sensitive? The phonological relationships between any given target word and its neighbors might be considered a relatively static influence on the target?s articulation ? for adult speakers of a language, the network structure of the lexicon is relatively stable.

However, speech production outside the laboratory occurs within the context of dynamic patterns of language use.

When speakers produce a word in context, how might that context modulate a neighbor?s influence on the target?s ultimate acoustic realization? One possibility is that a predictive semantic context (e.g., Superman tried to fly away but the phone booth door shut on his long red. . .) will boost the activation of a target (cape) relative to less contextually supported phonological competitors (e.g., gape, tape), or the context might reduce or restrict activation of words that are unlikely, semantically incongruent, or ungrammatical. In such a case, the influence of phonologically similar neighbors might be reduced (or even eliminated) in a highly semantically predictive context because unlikely neighbors are not competing as strongly. Alternatively, if contextual cues do not interfere with the dynamics of phonological neighborhood activation patterns during production processes, then neighborhood-conditioned influences on phonetic variation will remain strong irrespective of the context. That is, even if a target is highly predictable in context, the influence of the word?s phonological neighbors on its acoustic realization may persist.