The best of utterology

Well, I am not sure the title of this post is actually quite so accurate, but after such a long lag since my previous post I have to attract attention from my readers somehow!

In utterology-related developments around here, there's been lots of progress on lots of work. Most exciting at the moment are about 23,000 data points on an experiment I've just finished running in Jerusalem, thanks to my very hard-working colleague there. In this study, we'll hopefully be able to gain some insight into the effects of morphological family size in spoken word recognition. By replicating an earlier study by a different group of researchers who found a facilitatory effect of "related family size", or how many semantically related morphological relatives a word has, I hope to discover whether, on the one hand, the effect is mirrored in auditory word recognition, or whether, on the other hand, as with neighborhood density, the effect is reversed in the auditory domain.

One of the most amazing things that I keep remarking as I do experimental work is the amount of effort that must be poured into any investigation. In this particular experiment, I was quite fortunate to be able to simply use the very same stimuli (albeit in auditory form) that the original study used - I will be eternally grateful to the authors of that paper for providing me with those, because essentially I got a pre-packaged set of items that varied with respect to the relevant independent variables (frequency, related family size, and unrelated family size) of interest.

Next step involved finding a native speaker of Hebrew to patiently sit in the sound booth in our phonetics lab and record the stimuli. This was followed by careful choosing of the best exemplars, and then measuring their duration and, thanks to my electronic corpus of Hebrew, determining each stimulus's uniqueness point, since measuring reaction time to auditory stimuli can be a tricky and sometimes risky business. A good example is the first experiment I did on Maltese a few years back, where we got prematurely excited at finding an effect when measuring reaction time from stimulus onset, only to quickly realize that the relevant factor was strictly correlated with stimulus length - and indeed, the effect disappeared when reaction time was measured from stimulus offset.

In the current study on Hebrew, we have an even more accurate place to measure from: the lexical uniqueness point of each word. So now it is just a matter of finessing the data files a bit more, though to get them in the right shape I've had to ask a professional computational linguist to translate data files into the appropriate format readable by statistics software (luckily, I live with such a linguist, and luckily I had an excellent RA able to do a lot of the item-related measurements).

I may have exciting results to report soon, if the experiment worked as planned. But overall, my point here is that I continue to have deep appreciation of the efforts required to carry out experimental work.