Would you care for a beverage?

In the Sumner reading on “The social weight of spoken words”, I found the slend-er example interesting. I was able to connect this with a documentary I watched on English accents across different states and regions in America, where they actually showed that the accent which we prefer the most is the accent we hear most on television and media today. This relates to the fact mentioned in the paper that even people who themselves speak a stigmatized dialect tend to remember the “socially idealized” productions of words better. This also ties back to the lecture presented by Robert last week regarding the different dialects according to different geographic locations. I think that it is important to note that humans are predisposed to, often unconsciously, biases and discriminatory behavior earlier in process of spoken word recognition, so that we do not fall prey to these stereotypes when processing speech, and so that we can more effectively process speech in an effective way in our day-to-day lives.

In the King and Sumner reading, I found the study to be extremely important in improving upon existing models of spoken word recognition to include interaction between semantic association and speaker characteristics. The “exemplar resonance model” currently does not have the ability to discern differences in word association based on different social categories demonstrated based on phonetic cues. For instance, if I were to hear the word beverage spoken by someone with an American accent, I would think of coke, pepsi, and juice. But if I were to hear the word beverage spoken by someone with a British accent, I would think of tea, coffee, or hot chocolate. The fact that the two experiments in the study showed that individual words spoken by different speakers prompted different word associations did not surprise me at all, because I personally encounter examples of these biases in word associations based on the speaker. This ties back to the TICS Sumner reading, where the biases and stereotypes based on the speaker (judged based on the voice for gender, age, status, race, etc.) affected how we process and recognize spoken words.

In the Sumner and Kataoka reading, I was interested in seeing the metrics on recall of words based on different speaker accents. This was a concept that was introduced in the TICS paper, for the slend-er example, and I was surprised to learn that the recall ability of GA listeners for GA vs. BE speakers were extremely similar. I would have expected people who are GA listeners to be primed and predisposed to processing and recalling GA speaker better than BE speakers since GA is the speaker accent that GA listeners are most familiar with.

All three papers shed light on how we perform spoken word recognition based on many social characteristics like dialect and gender. All three papers are a call to the linguistics field for improved models that properly incorporate these information sources into the model spoken language understanding.