Day 68-69

Saturday! The weekend. It started off by magically finding my phone. That would be the story for the rest of my weekend. Anyway, Saturday I worked a little bit on a project until later that night. Come night fall, Myself, Benny, Dave, Rob, Jacob and Marc headed to the pool hall. It was a good 30 minute walk away, but it was well worth it. We were able to get two tables for two hours, for a total of 8 pounds. Not bad if you ask me. Definitely a fun night, haven't shot pool for a while.

Today was even better. It started off with an early wake up time and an 11:45 tee time at Braid Hills, one of Scotlands Top 100 golf courses, and one of Edinburgh's hardest. Shahab Marc and I arrived just around 11:40, plenty of time to not warm up at all. . . When we got to the proshop he informed us that all his computers, phones, and card machines are down - meaning our credit/debit cards won't work! So, he proposed that on hole number 5, we take a detour to a neighboring and connected course and pay there. But by the time everything was sorted out and we got the rental clubs, the machines came back online! Yet he couldn't sign in to the till etc... so we left our cards there and just started our round.

It was an absolutely beautiful course. With panoramic views of all of Edinburgh. It was also one of the most hilly courses I've seen in a while, if not ever. Luckily, it wasn't the longest course, which made the blind tee shots and tough holes accessible via irons. I played exceptionally well for not having played in a while, ended up shooting 3 over. Highlights of the day were a 410 yard drive and another drive that was about 300 up hill which flew over a guys head who was playing that hole landed right in front of the hole (picture), of which I got my only birdie >.<

Some of the landscape of the course...

The up hill drive!

A great panoramic shot of Edinburgh...

Another great view

Me standing on the 18th tee...

After golf, we headed to Shahab's place, he made some pasta that was damn good.

When I got back home I started out on one of the two projects I have left to do.

This one is on the Cohort Model of Speech Perception...

What the general consensus in speech perception research is at the moment, is that there is competition in our lexicons (essentially a giant word-bank in our mind) between similar sounding words. The Cohort Model states that acoustic information required to identify a word is revealed over time. The central idea of the model is that a large number of words, in parallel, are considered as "candidates" and gradually decay away when more evidence is provided (more of the word, context, etc). For example, when the sound "e" is presented, all words beginning with the "e" sound are accessed, this is called the "cohort" of words. When more information is provided, such as "ele", the cohort gets smaller (elephant, electricity, etc). This proceeds until a point is reached, "eleph", where the evidence is consistent with only one word, elephant, this is deemed the "uniqueness point". With this being said, the start of the word, and more specifically the first syllable, is extremely important. This is shown through research, one experiment shows that listeners are better at detecting speech distortions prior to the uniqueness point. Going with this further, in a study on "fluent restorations", in which subjects made "corrections" when repeating speech that had distortions in it, it was shown that more restorations were made if the distortion was: slight, at the end of the word, or in predictable context. Further supporting sensitivity to beginning of words. Another study showed that the time it takes to recognize a word depends on when the uniqueness point is, however, context effects will allow selection before the uniqueness point. Granted, this can lead to possible errors. So, what about context? What about word frequency (how often you see a word)? In a study on word frequency, it was found that not all candidates are equivalent. For example, "spee" is more likely to activate high frequency words such as speech and speed before low frequency words such as species. This suggests that word frequency has an early effect in speech recognition. Now, what about the semantic context? In initial versions by Marslen-Wilson & Tyler (the cohort models parents), they state that context could eliminate items in the cohort. Now, with new studies being presented, this has changed and it is said in the later versions of the model that context is only involved in the interaction stage of speech production - in which we use both semantic context and word frequency to integrate the word into a sentence and a mental model. The evidence for this is seen through numerous different experiments. First, reaction time studies have shown that perceptually ambiguous speech activates the full cohort, even contextually inappropriate words. More evidence, is the Gating Task, in which increasing amounts of a word was shown (a...ap...app...appl...apple), when no context was reported - the average response was 333msec, when context was provided, the response time dropped to 199msec. The last evidence supporting late effect of context in speech recognition is that of electrophysiology studies. More specifically N400 tests, which means there is a negative peak of electrical current around the 350-400msec mark, this peak appears to be particularly sensitive to semantic anomalies. In one of these studies, there was a N400 found with semantically abnormal sentences, such as, "The Dutch trains are sour" - obviously, a train can't be sour. The N400 was also found during semantically conflicting sentences, such as, "The Dutch trains are yellow" (when in fact they are white; the subjects knew this). In another N400 study, the researcher presented homologous word triplets, such as "River-Bank-Money" and "Finance-Bank-Money"... the N400 was found in the former but not the latter. This suggests that the N400 reflects global context versus local context, due to that fact that it operates over all three words - given the last two words are the same in both conditions. And that my friends, is the Cohort Model of Speech Perception.