Loading wiki pages...

Wiki Version:
<p><em>Syntactic adaptation effects do not transfer across tasks</em></p> <p>Emily Atkinson,1 Ian Rigby,2 Naomi Shapiro,2 Brent Woo,2 & Akira Omaki2</p> <p>1 University of Michigan; 2 University of Washington</p> <p>Studies on syntactic adaptation have demonstrated that biases in ambiguity resolution can be altered by manipulating the probability of the competing structures in the input [1-3]. A recent experiment extended these findings to filler-gap dependency processing [4]. The parser is biased to complete filler-gap dependencies early (i.e., at the verb) [5] in part because direct object (DO) gaps are the most frequent gap position [4]. Exposure to input skewed toward an <em>a priori</em> unexpected gap position, i.e., prepositional object (PO) gaps (1:<em>wrote the post about __</em>), alters this bias such that participants no longer actively complete the dependency at the verb. However, these existing demonstrations come from studies that manipulate the distribution of structures within a single task. In these instances, adaptation may be the result of task-specific expectation adaptation rather than a general adaptation of parsing biases. If adaptation effects transfer across different tasks (as in phonetic adaptation studies [6]), this would provide evidence that syntactic adaptation is the result of changes to underlying parsing biases. While previous studies have utilized blocked designs to provide concentrated input [2,4], the current studies present this exposure as a separate task. Two eye tracking during reading experiments provide evidence that syntactic adaptation effects do not transfer from one task environment to another.</p> <pre class="highlight"><code> In both experiments, the exposure phase was masked as a</code></pre> <p>separate sentence recognition study: participants read 12 short stories that each contained 4 target sentences and were tasked with identifying if 2 sentences were contained in the previous story. For one group of participants, all of the target sentences contained DO gaps (1:<em>wrote <strong> about the article<em>), while the other group’s stories contained PO gap target sentences (1:</em>wrote the post about </strong></em>).</p> <pre class="highlight"><code> *Exp1 *(N=48) Target sentences in both the story phase and test</code></pre> <p>phase consisted of a filler-gap dependency with a PO gap preceded by a direct object NP (2). If the parser actively completes the dependency at the verb, we expect a <em>filled gap effect</em> [5,7] on the DO region: a reading time increase in the NP-fronting condition (2:<em>that</em>) compared to the PP-fronting condition (2:<em>from which</em>). In the critical DO region, the NP-fronting condition had longer regression paths (<em>β</em>=102.42, <em>t</em>=4.90, <em>p</em>&lt;0.001) and more regressions (<em>β</em>=0.62, <em>Z</em>=3.02, <em>p</em>&lt;0.01) than the PP-fronting condition. There was no significant effect of story exposure group or an interaction. Thus, exposure group did not affect filler-gap dependency processing during the test phase as participants in both groups exhibited the filled gap effect.</p> <pre class="highlight"><code> *Exp2* (N=48) While the story phase was identical to that from</code></pre> <p>Exp1, the target sentences during the test phase utilized a <em>plausibility mismatch </em>manipulation [8] to examine active gap filling. Specifically, we expect a slowdown on or after the verb in the implausible-filler conditions (3:<em>which city…wrote</em>) compared to the plausible-filler conditions (3:<em>which book</em>…<em>wrote</em>). No significant effects were found in the critical verb region, but results from the spillover region indicate that the plausibility manipulation was successful. Sentences with implausible fillers had significantly longer regression path durations (<em>β</em>=-38.24, <em>t</em>=17.95, <em>p</em>&lt;0.05) and significantly more regressions (<em>β</em>=-0.55, <em>Z</em>=0.18, <em>p</em>&lt;0.01). The effect of story exposure group and the interaction were both not significant. As in Exp1, exposure group did not affect dependency processing.</p> <pre class="highlight"><code> Taken together, these results suggest that syntactic adaptation is</code></pre> <p>not the result of generalized changes to parsing biases, but rather a more specific shift in parsing biases that may be task-specific. Implications for the theory that syntactic adaptation is a form of implicit learning [3] will be discussed.</p> <p>(1) <em>DO gap exposure / PO gap exposure:</em> The post that their friend wrote <strong> about the article… / The article that their friend wrote the post about </strong>...</p> <p>(2) <em>PP-fronting / NP-fronting:</em> The suitcase <em>from which</em> / <em>that</em> the stealthy, wanted thief |V stole |DO the precious jewels | <strong> / <em>from</em> </strong> was full of sentimental items.</p> <p>(3) <em>Plausibility Match / Mismatch:</em> <em>The book</em> / <em>The city</em> that |NP the author |V wrote |Spillover thoughtfully | about __ was named for an explorer.</p> <p><em>References</em> [1] Jaeger & Snider 2013. <em>Cognition</em>. [2] Fine et al. 2013. <em>PLOS One</em>. [3] Fine & Jaeger 2013. <em>CogSci</em>. [4] Atkinson & Omaki 2016. CUNY poster. [5] Stowe 1986. <em>LCP</em>. [6] Norris, McQueen, & Cutler 2003. <em>Cognitive Psychology</em>. [7] Wagers & Phillips, 2014. <em>JEP</em>. [8] Traxler & Pickering 1996. <em>JML</em>.</p>
OSF does not support the use of Internet Explorer. For optimal performance, please switch to another browser.
This website relies on cookies to help provide a better user experience. By clicking Accept or continuing to use the site, you agree. For more information, see our Privacy Policy and information on cookie use.

Start managing your projects on the OSF today.

Free and easy to use, the Open Science Framework supports the entire research lifecycle: planning, execution, reporting, archiving, and discovery.