Download the full-sized PDF of Medial stop reduction and word recognition “in the wild”Download the full-sized PDF



Permanent link (DOI):


Export to: EndNote  |  Zotero  |  Mendeley


This file is in the following communities:

Linguistics, Department of


This file is in the following collections:

Research Publications (Linguistics)

Medial stop reduction and word recognition “in the wild” Open Access


Author or creator
Tucker, Benjamin V.
Brenner, Daniel
Sims, Michelle
Additional contributors
spontaneous speech
word-medial stops
Type of item
Conference/workshop Presentation
Research on laboratory speech has uncovered many aspects of speech perception and comprehension, however we still know very little about the recognition of the more common spontaneous types of speech. This presentation reports an experiment designed to investigate the relationship between word-medial stop production variability and word recognition. Specifically, this experiment tests whether there is a relationship between the production frequency and the range of variation in the production of word medial stops which directly affects recognition. In spontaneous speech, speakers often produce “reduced” speech forms: for example, a phrase like “Did you eat yet?” is often produced as [dʒitʔjɛtʔ] or “jeet yet?” Reductions are extremely common in everyday speech. One study of conversational speech found that 25% of word tokens differ from their dictionary pronunciation (Dilts, 2013). Research on this topic has repeatedly shown that reduced speech is more difficult to process than unreduced speech (e.g., Ernestus et al., 2002; Tucker, 2011; Ven et al., 2011). These studies generally have dichotomized reduction into two groups, reduced and unreduced. Tucker (2011) also found in a post-hoc analysis of the lexical decision latencies that by using an acoustic measure to represent the range of reductions that a possible non-linear relationship emerges between reduction and response latencies. This non-linear relationship may indicate that words produced with their most frequent pronunciation variation elicit the fastest responses, however the stimuli used in Tucker (2011) were not designed to specifically investigate a correlation between a gradient acoustic measure, intensity difference, and response latency. It is also possible that the unreduced form is the lexically stored form and that when using a gradient acoustic measure the more deviant the input is from that unreduced form the slower the response time will be resulting in a more linear type of response. The present study, then, directly tests the distributional hypothesis implied by Tucker (2011) by using a large variety of word-medial stops, which cover the full production spectrum (reduced, unreduced, and everything in between). Items were selected from a corpus of spontaneous speech allowing for access to a much broader array of stop production compared to the dichotomous distribution used in Tucker (2011). In following part of Ernestus et al. (2002)’s methodology, we extracted words containing word-medial stops in three different contextual environments: “isolation”, “phonological”, and “phrasal” contexts. Unlike the methodology used in Ernestus et al. (2002), we conducted a cross-modal identity priming experiment, which allows for a more direct investigation into processing load (and yields a more refined measure than a word identification task). Seventeen native English listeners have participated in the experiment, with each participant responded to a total of 366 items (54 identity primes, 54 controls, 60 with phonological overlap, 36 “yes” distractors and 162 non-words). Initial results indicate that some degree of context surrounding the target word facilitates recognition (e.g. the phonological context), however in the full context, responses were slower than the other contexts, likely due to the increased amount of lexical competition by the introduction of more words. Overall we find support for previous work: generally, more reduced items are more difficult to process. It is notable that unlike Tucker (2011), we do not find any non-linear relationship between our reduction measure (intensity difference) and response time, but we find a linear relationship which we interpret as indicating that it is indeed the unreduced form that is stored as opposed to the more distributionally valid form of the word. Tucker, B. V., Brenner, D., Sims, M. N. (2014). Medial stop reduction and word recognition “in the wild”. The Ninth International Conference on the Mental Lexicon. Niagara-on-the-Lake, ON.
Date created
License information
Creative Commons Attribution 3.0 Unported

Citation for previous publication

Link to related item

File Details

Date Uploaded
Date Modified
Audit Status
Audits have not yet been run on this file.
File format: pdf (Portable Document Format)
Mime type: application/pdf
File size: 1597522
Last modified: 2015:10:12 19:48:16-06:00
Filename: ML_Presentation_2014-10-01.2.pdf
Original checksum: 265b26e844d49bb5f542a6e27f32df69
Well formed: false
Valid: false
Status message: Annotation dictionary missing required type (S) entry offset=1072198
File title: 1. Introduction
File title: Medial stop reduction and word recognition ``in the wild''
File author: Benjamin V. Tucker*, Daniel Brenner**, Michelle Sims*
Page count: 23
Activity of users you follow
User Activity Date