Individual differences in prediction: An investigation of the N400 in word-pair semantic priming Xiao Yang & Lauren Covey Cognitive and Brain Sciences Brown Bag Talk October 17, 2016
Caitlin Coughlin, PhD Robert Fiorentino, PhD Cynthia Siew Maria Teresa Martinez-Garcia, PhD Adrienne Johnson, PhD Data collected in the KU Neurolinguistics and Language Processing Lab 2
Research Questions During comprehension, do individuals make predictions about upcoming lexical material? If so, what makes one a better predictor? 3
Anticipating words in context Many studies have used semantically supportive contexts to investigate brain responses to a highly predictable word She saw a dog chasing a cat In discourse- or sentence-level experiments, it is difficult to tease apart whether anticipatory effects reflect active prediction, or whether the supportive semantic context helps to ease integration of the target word (e.g., cat) 4
Word-pair semantic priming (prime) (target) CAP hat Related KEY hat Unrelated After encountering a prime, it is easier to process a semantically related target, than an unrelated one. 5
Using EEG to study priming effects EEG (electroencephalography) Records real-time voltage fluctuations in the brain Captures brain activity triggered by a particular linguistic event (e.g., encountering a related or unrelated word) 6
ERP component of interest: N400 N400: A negative-going wave peaking at 300-500ms N400 priming effect: Related word pair elicits N400 reduction in amplitude (Lau et al., 2013) CAP hat Unrelated KEY hat Related (Negative plotted up ) 7
N400 priming effect What is the priming effect due to? Mechanism 1: Passive association The activation of the prime word passively spreads to semantically related words in the mental dictionary The N400 priming effect is based solely on the semantic relatedness between the prime and the target 8
N400 priming effect What is the priming effect due to? Mechanism 2: Active prediction The priming is in part due to an active prediction process in which the individual generates an expectation for a specific word that is semantically related The N400 priming effect for related words is, at least in part, due to an effect of a confirmed prediction 9
Mechanism 1: Passive association CAP HAT KEY
Mechanism 1: Passive association cap CAP HAT KEY
Mechanism 1: Passive association cap CAP HAT KEY
Mechanism 1: Passive association cap CAP HAT KEY
Mechanism 2: Active prediction CAP HAT KEY
Mechanism 2: Active prediction cap CAP HAT KEY
Mechanism 2: Active prediction cap CAP HAT KEY
Mechanism 2: Active prediction cap Since cap just appeared, I bet hat will appear soon KEY CAP HAT
Lau et al. (2013) Utilize word-pair semantic priming in lieu of sentence context manipulations so that the strength of the semantic associations between primes and targets is held constant Related Trials Prime Target DILL pickle JIGSAW puzzle DAY night Forward association strength of 0.5 and a mean forward association strength =.65. South Florida Association Norms (Nelson, McEvoy, & Schreiber, 2004) 18
Lau et al. (2013) While semantic relatedness is held constant for Related and Unrelated pairs, a block design was implemented to test whether predictive validity would modulate the N400 effect Block 1 20% Related Prime-Target Pairs Block 2 80% Related Prime-Target Pairs Behavioral studies report increased priming when there are a high proportion of related trials (e.g., Hutchison, 2007) 19
Lau et al. (2013): Predictions Passive association account: N400 priming effect expected for Related-Unrelated comparison Relatedness proportion is not expected to influence N400 priming effect Active prediction account: N400 priming effect expected for Related-Unrelated comparison Greater priming effect expected for high-relatedness block than for low-relatedness block 20
Lau et al. (2013): Results N400 priming effect in both low- and highrelatedness blocks N400 effect was larger in high-relatedness block Lau et al. (2013) argue in favor of an active prediction account 21
The present study Builds on the methodology and design of Lau et al. (2013) by incorporating an experimental manipulation from the behavioral literature Relatedness proportion in the present study is fully randomized on a trial-by-trial basis, following Hutchison (2007) Relatedness Cue: modulates reaction times during word-pair priming (Hutchison, 2007) 22
Relatedness Cue For a given trial, a relatedness cue is presented before the word pair, indicating how likely it is to encounter a related word pair (proportional cue reflects real proportions) Cue (1000ms) Prime (500ms) Target (900ms) 80% Related CAP hat 20% Related KEY hat 23
Experiment design: Conditions Related Condition Unrelated Condition Relatedness Cue Prime Target Prime Target 80% Related CAP hat KEY hat 20% Related CAP hat KEY hat 160 prime-target word pairs 280 filler pairs, used to establish relatedness proportion Task: semantic probe detection (animal words) 24
Prediction & Cognitive Abilities If individuals engage in predictive processing, then individual differences in the cognitive abilities that underlie prediction may modulate the effect. Attentional Control (including Working Memory): maintaining attention; retrieving correct information during interference; inhibiting habitual responses (Kane & Engle, 2002; Daneman & Carpenter, 1980; Just & Carpenter, 1992) Verbal Fluency: the ability to quickly and accurately generate words in response to a given cue (Federmeier et al., 2002; Federmeier, Kutas, & Schul, 2010) 25
Attentional Control (AC) Hutchison (2007): only high-ac participants showed an effect of relatedness-proportion on the magnitude of the priming effect Effortful process of generating predictions only when it is most beneficial to do so (i.e., in the high-relatedness condition) requires attentional control We assess Attentional Control via a number Stroop task: one one one 1 2 3 4 26
Working Memory (WM) Several studies have implicated a role for working memory in the recruitment of semantic information during processing, which is potentially very important during prediction (e.g., Van Petten et al., 1997; Nakano et al., 2010) We assess WM via a Counting Span task (Conway et al., 2005) 27
Verbal Fluency (VF) Verbal fluency affects the extent to which older adults are sensitive to semantically unexpected words (e.g., DeLong et al., 2012; Federmeier et al., 2002; Federmeier et al., 2010) We assess VF via an offline letter and category task (Benton & Hamsher, 1978) Within 1 minute, tell me all the words for office supplies that you can think of. Paper, Stapler, Eraser, Sticky notes, Pen 28
Experimental procedures Participants: 22 right-handed native English speakers Procedures: Offline measures counterbalanced in presentation order Letter and category task (VF) Stroop (AC) Count Span (AC) EEG experiment 29
Research Questions & Predictions 1. What is the effect of Relatedness? Related targets should elicit attenuated N400 compared to Unrelated targets 2. What is the effect of Relatedness Proportion? Related targets in 80% condition should show greater N400 attenuation compared to Related targets in 20% condition (e.g., Hutchison, 2007; Lau et al., 2013) 3. Will attentional control or verbal fluency modulated the N400 effect? If ERP evidence implicates active prediction process, individual differences should modulate the N400 effect 30
Statistical analyses EEG data was analyzed using linear mixed-effects models with the lme4 package in R Time window of investigation was 300-500 ms Fixed factors: Relatedness (Related, Unrelated) Cue (80%, 20%) Verbal Fluency (Letter + Category score) Composite Attentional Control (Stroop + Count Span) Hemisphere (Left, Midline, Right) Anteriority (Anterior, Central, Posterior) Random intercept: Subjects 31
ERP Results: Relatedness Related targets exhibit N400 attenuation compared to Unrelated targets 80% Related Condition 20% Related Condition 32
ERP Results: Verbal Fluency In the high-relatedness condition ( 80% ), individuals with greater verbal fluency show greater attenuation of the N400 than individuals with lower verbal fluency Low-VF Participants 80% Related Condition High-VF Participants 80% Related Condition 33
ERP Results: Attentional Control In the low-relatedness condition ( 20% ), individuals with greater attentional control scores show less attenuation of the N400 than individuals with lower attentional control Low-AC Participants 20% Related Condition High-AC Participants 20% Related Condition 34
Summary of Results 1. Related targets yielded attenuated N400s compared to Unrelated targets 2. Effect of Relatedness Proportion (i.e., Cue) was differentially modulated by measures of verbal fluency and attentional control Verbal Fluency: In high-relatedness condition (e.g., 80% ), higher verbal fluency resulted in greater N400 attenuation Attentional Control: In low-relatedness condition (e.g., 20% ), higher attentional control resulted in less N400 attenuation 35
Summary of Results Verbal Fluency: In high-relatedness condition (i.e., 80% ), higher verbal fluency resulted in greater N400 attenuation The ability to quickly generate a set of semantically/ phonemically related words has implications for predictive abilities (e.g., DeLong et al., 2012; Federmeier et al., 2002; Federmeier et al., 2010) Verbal Fluency and Prediction: 1. Production (covert) 2. Inhibition of irrelevant lexical items 3. Vocabulary size 36
Summary of Results Attentional Control: In low-relatedness condition (i.e., 20% ), higher attentional control resulted in less N400 attenuation Individuals with greater attentional control have a greater ability to inhibit lexical prediction when it is unproductive for them to do so (e.g., Hutchison, 2007) 37
Conclusions Provides converging evidence in support of Lau et al. (2013) for an active prediction process in the N400 Finds effects of predictive validity on a trial-by-trial basis, removing potential concerns related to block design Builds on Hutchison (2007) to implicate a role for verbal fluency and attentional control in predictive abilities Provides electrophysiological evidence by using the N400 effect One of the first studies to find a relationship between VF and predictive processing in young adults 38
Thank you! Project team members: Caitlin Coughlin, María Teresa Martínez García, Adrienne Johnson, Cynthia S. Q. Siew, and Robert Fiorentino, and Spring 2014 Neurolinguistics II class members References: Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), 1-48. doi:http://dx.doi.org/10.18637/jss.v067.i01 Benton, A., & Hamsher, K. (1978). Multilingual aphasia examination manual. Iowa City, IA: University of Iowa. Conway, A., Kane, M., Bunting, M., Hambrick, D., Wilhelm, O., Engle, R. (2005). Working memory span tasks: A methodological review and user's guide. Psychonomic Bulletin and Review, 12(5), 769-786. Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of verbal learning and verbal behavior, 19(4), 450-466. DeLong, K. A., Groppe, D. M., Urbach, T. P., & Kutas, M. (2012) Thinking ahead or not? Natural aging and anticipation during reading. Brain & Language, 121, 226-239. Federmeier, K. D., Kutas, M., & Schul, R. (2010). Age-related and individual differences in the use of prediction during language comprehension. Brain and language, 115(3), 149-161. Federmeier, K. D., McLennan, D. B., Ochoa, E., & Kutas, M. (2002). The impact of semantic memory organization and sentence context information on spoken language processing by younger and older adults: An ERP study. Psychophysiology, 39(2), 133-146. Hutchison, K. A. (2007). Attentional control and the relatedness proportion effect in semantic priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(4), 645. Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehension: individual differences in working memory. Psychological review, 99(1), 122. Kane, M. J., & Engle, R. W. (2002). The role of prefrontal cortex in working-memory capacity, executive attention, and general fluid intelligence: An individual-differences perspective. Psychonomic bulletin & review, 9(4), 637-671. Lau, E. F., Holcomb, P. J., & Kuperberg, G. R. (2013). Dissociating N400 effects of prediction from association in single-word contexts. Journal of Cognitive Neuroscience, 25(3), 484-502. Nakano, H., Saron, C., & Swaab, T. Y. (2010). Speech and span: Working memory capacity impacts the use of animacy but not of world knowledge during spoken sentence comprehension. Journal of Cognitive Neuroscience, 22(12), 2886-2898. Van Petten, C., Weckerly, J., McIsaac, H. K., & Kutas, M. (1997). Working memory capacity dissociates lexical and sentential context effects. Psychological Science, 8(3), 238-242. 39
Appendix I: Cognitive Scores Verbal Fluency Score Stroop Accuracy Effect Stroop RT Effect Count Span Accuracy Mean 86.591-2.955-73.563 68.748 StDev 31.921 5.111 69.020 13.806 Minimum 38.000-15.000-279.514 41.440 Maximum 172.000 7.500 56.197 88.000 40
AC z-scored Appendix II: FAS/AC correlation 3 2 1 0-2 -1.5-1 -0.5 0 0.5 1 1.5 2 2.5 3-1 -2 FAS z-scored 41
Appendix III: Final Model(s) Where are the effects sig? (anteriority/laterality) 42