“The Added Value of Musical Training on Linguistic Syntax Processing”
presented with acclaim at:
• ASHA Convention, Orange County Convention Center, Orlando, FL, November 2019
• Cognitively-Based Music Informatics Research (CogMIR) Seminar, Tow Center for the Performing Arts, Brooklyn, NY, August 2019
• Evidence-Based Research Poster Day Presentation, Hunter College Brookdale Campus, New York, NY, April 2019
Main Argument
Connections between linguistic and musical syntax is a growing topic of interest. Both involve complex sequences that unfold rapidly in time and share key properties of generativity, abstract relations, and hierarchy. Brain damage may interrupt language and music processing (aphasia and amusia, respectively) suggesting separate underlying mechanisms. However, neuroimaging reveals overlapping neural correlates that contradict modularity. As a compromise, the shared syntactic integration resource hypothesis (SSIRH) distinguishes between domain-specific representations in long-term memory and overlapping resource networks that activate these representations during processing. In support of SSIRH, Slevc, Rosenberg, and Patel (2009) demonstrated when unexpected linguistic structures (e.g., garden path sentences) are simultaneously paired with unexpected harmonic structures (e.g., out-of-key chords) during self-paced reading, reading times (RTs) in participants exponentially increase. However, this research did not control for musical proficiency. Given a growing body of evidence supporting musical training enhancing a variety of linguistic skills, inclusion of this group is important because it allows for examination of the added value of musical training in sequenced language processing. This study investigated if differences exist in linguistic and musical syntax processing between musicians and nonmusicians by replicating procedures by Slevc, et al. (2009).
Procedures
Forty monolingual individuals, ages 28-44, participated. Twenty musicians received a minimum of 11 years musical instruction; seventeen of whom received the equivalent of a four-year baccalaureate degree in musicology or performance from a formal conservatory to ensure proficiency. Twenty nonmusicians received a four-year baccalaureate degree outside of the musical arts and never proficiently learned an instrument nor received long-term musical education.
Participants completed self-paced reading whereby they read sentences by pressing a button to present consecutive text segments on a computer screen. A musical chord heard over headphones accompanied each segment. After each sentence, participants answered a yes/no comprehension question. Participants were instructed to quickly read the sentences, but remaining attentive enough to answer the questions correctly. Items were presented in a pseudorandom order, such that no more than two consecutive trials contained out-of-key chords. Two within-participants factors were included in this experimental design: linguistic expectancy and musical expectancy.
Stimuli
Within 12 critical trials, syntactic expectancy was manipulated by including a garden path sentence, resulting in an unexpected syntactic interpretation at the critical region (underlined in Example 1, below). Thirty-six additional filler trials were included that contained no unexpected elements (see Example 2, below). Thus, only 25% of total sentences contained garden path sentences, making the conscious perception of the manipulations unlikely.
(1) The teacher recalled the answer to the question was written in the book.
(2) The forecast predicted rain, but it was a sunny day.
Each sentence was accompanied by a uniquely composed chord sequence. The length of these chorales ranged from eight to 11 chords and occurred at least five chords into the sequence to establish key. Of the 48 total chorales, 36 contained all chords adhering to the single key which paired with the filler trials. The 12 paired with the critical linguistic items were identical except for the substitution of 1 chord. This substitution contained a distantly-related key three, four, or five keys away and corresponded to the syntactically unexpected region of the critical sentence (e.g. was, above).
Raw data was recorded using Psychopy software. RTs and accuracy data were uploaded and analyzed in Microsoft Excel. Pre-critical, critical, and post-critical values were calculated and differences between RT means were compared with regression analysis and t-test statistics set to 5% (a=0.05) significance levels and 95% confidence intervals.
Results
Participants showed comparable RTs in pre-critical regions throughout the trials. However, musicians averaged marginally faster RTs than nonmusicians in these regions which support previous research pointing to reading skills differing with musical ability (musicians: `x = 0.658, s = 0.342, E = 0.433; nonmusicians: `x = 0.706, s = 0.366, E = 0.046, RT difference = 0.048). These data are supported with regression analysis and t-testing (b = -0.006, SE = 0.349, t = 1.49, p = 0.137). Within critical regions, nonmusicians showed substantially enhanced garden path effects as compared to musicians (musicians: `x = 0.669, s = 0.339, E = 0.042; nonmusicians: `x = 0.824, s = 0.433, E = 0.054, RT difference = 0.155). These data reflect a main effect of expectancy (b = 0.147, SE = 0.395, t = 4.36, p = 0.0000161). Post-critical regions showed a continued difference in RTs between participants: RTs were substantially slower in nonmusicians while RTs in musicians were observed to be relatively unaffected (musicians: `x = 0.634, s = 0.323, E = 0.041; nonmusicians: `x = 0.781, s = 0.418, E = 0.529, RT difference = 0.147). These data also reflect a main effect of expectancy (b = 0.0537, SE = 0.262, t = 4.309, p = 0.0000202). Interestingly, musician RTs in these post-critical regions were faster as compared to their pre-critical regions. Analyses of post-sentence comprehension questions between filler trials and critical trials were also notable. Musicians achieved overall lower accuracy percentages than nonmusicians on the filler questions (musicians: `x = 90.6, s = 2.3; nonmusicians: `x = 92.3, s = 2.1, difference = 1.7%). However, nonmusicians achieved comparatively far lower accuracy percentages than musicians on the critical questions (musicians: `x = 90.1, s = 2.2; nonmusicians: `x = 85.9, s = 2.0, difference = 4.2%). This suggests interpretation difficulties were present during critical trials in nonmusicians as compared to musicians, a possible further reflection of the competing nature of syntactic resource for language and music.
Discussion:
The present experiment provided evidence for shared processing between mechanisms of linguistic and musical syntax, demonstrated through simultaneous manipulations present during difficult integrations. These observations support research suggesting resources for processing incoming linguistic and musical syntax are finite and competing. Importantly, musicians in this research were not slowed during, or after, critical regions of expectancy, pointing to linguistic syntax processing positively differing with musical ability. These data provide a new opportunity to better understand the implicit benefit of musical training on a mechanistic level, and argue for an improvement in quality and quantity of music training across the lifespan.
full thesis and list of references provided upon request.