Anglicizing Spanish /a/

Implications of statistical distribution and language experience on non-native pronunciation

Senior Honors Thesis by Kira Prentice

BACKGROUND

BACKGROUND

Spanish vowel /a/ into English /æ/ and /ɑ/

What else conditions the nativization?

MAIN QUESTIONS

• How do English speakers nativize the vowel /a/ in words that they perceive to be from Spanish?
• HYPOTHESIS? Statistical calculations of /ɑ/ and /æ/ frequencies as a function of immediate consonantal context in English
• Across factors:
  • With and without Spanish experience?
  • Perceived source language?

GENERATING A HYPOTHESIS

Nonce words that are orthographically and phonologically legal in both English and Spanish

• Disyllabic CVCV structure, with "a" in the stressed syllable
• Eliminate ambiguous consonantal context across both languages
• Occurrence of /ɑ/ and /æ/ vowels in these contexts

GENERATING THE FREQUENCIES

Using CMUDict, an online pronunciation dictionary

def context(token):

  # Open file to read
  words = open('./cmudict', 'r')

  # Create output file
  file_name = "./output/output_" + token + ".csv"
  output = open(file_name, 'w')

  # Allowed phones in output
  filter_pre = ['T', 'K', 'B', 'S', 'P', 'D', 'F', 'CH', 'G'];
  filter_post = ['T', 'K', 'B', 'S', 'P', 'D', 'F', 'CH', 'G', 'M', 'N'];

  for line in words:
    if token in line: 
	  # Truncate to surrounding context
	  # Check if token is first or last in word
	  # Make sure surrounding context is included in the filter
	  output_string = s[0] + ',  ,' + s[ndx] + ', ' + s[ndx+1] + '\n'

	  output.write(output_string)

  output.close()
  words.close()
					

This thesis is Open Source! Code available at github.com/kcp288/senior-thesis.

SAMPLE WORD LIST

DESIGN OF STUDY

• 22 participants total:
  • 10 with Spanish experience
  • 12 without Spanish experience
• Age range 18–32 years old
• All native U.S. English speakers
• All recruited from NYU
• Average Spanish experience: 8 years

PROCEDURE

• Two randomized word lists, with a total of 79 items
• Told one list was "ENGLISH" and one was "SPANISH"
• Read each word off slides, self-paced

RESULTS

• Source language of the token had a significant effect on phone produced
  • "a" as /ɑ/ when told it came from Spanish
• Participant language background also had significant interaction with main effect
  • "a" as /ɑ/ by participants with Spanish experience
  • Spanish group (with Spanish-primed tokens), used /æ/ only 4% of productions
  • Versus non-Spanish group (with Spanish-primed tokens), used /æ/ for 37% of productions

RESULTS

RESULTS

• Any significant effects of context failed to converge, including:
  • Examination of place (initial, coronal, labial or velar)
  • Examination of manner (initial, obstruent or nasal)
  • CVCV token structure?

RESULTS

• Minor effects of context
  • Vowel-initial environments favor /æ/
    • "Actu", "Aplo"
  • Higher than average appearance of /æ/ with velar-initial (/k/) consonants
    • "Casu", "Cano"

RESULTS

SUMMARY

• Even without Spanish experience, speakers have a concept of "what Spanish sounds like"
• Language experience affects production of Spanish versus English, perhaps influencing this underlying concept
• Further investigation of context effects

FURTHER QUESTIONS

• Using acoustic data; more subtle shifts
• Work with existing loanwords, running speech
• Beyond immediate consonantal context
• Sociolinguistics effects

WHY IS THIS IMPORTANT?

QUESTIONS?

Anglicizing Spanish /a/: Implications of statistical distribution and language experience on non-native pronunciationSenior Honors Thesis by Kira Prentice

Find out more at github.com/kcp288/senior-thesis.