• I extended the searches back to January 12th. (I originally began on January 26th, the day after the New Hampshire debate which featured Peter Jennings' question to Wesley Clark about Michael Moore and his "deserter" charge.)



• I also fine-tuned the searches a little. The four search expressions I used were:
  
  
  1. Bush deserter -AWOL (stories which had "deserter" only, and not "AWOL")
  
  
  2. Bush AWOL -deserter ("AWOL" but not "deserter"
  
  
  3. Bush +AWOL +deserter (stories which used both terms)
  
  
  4. Bush +"service record" -deserter -AWOL (stories which used "service record" but did not have either of the other terms)
  
  (I also tried searching on "military record" but there were too many stories to begin with, too much noise drowning out the effects I has looking for. There was noise in the other counts as well -- for instance, stories about Paul O'Neill being AWOL in Africa at the time of a stock market crisis -- but not nearly as much.)



• I recorded the counts of these 4 searches for each date between 1/12 and 2/8, and then totaled each day



• I then graphed the result, but chose to start the graph at 1/23, when the "deserter" story started to take off.



• It should be noted that Google News keeps stories current for 1 month, so older stories are slipping away, reducing each count. This will start to be a factor by the middle of this month, when significant number of "deserter" stories start disappearing, and near the end of the month when the pace picks up considerably.



• Because of this, I also counted and graphed each day's "adds", the number of stories added in each category that day, and the total number of adds for the day. On this graph it isnt quite as easy to see the trajectory of things, but it will stay accurate longer than the "stories" graph.



• One further note, the numbers I've been posting were all collected around the middle of the day on the date noted (usually between 12n and 5pm, sometimes later in the evening), because that's when I originally thought of doing it, and I continued that practice once I started. On this new data set, since I used Google News' search parameters to limit the days being searched, each day's numbers represent all the stories dated that day. This means that a graph of the numbers I posted previously would look a little different, but the shape of things would be just about the same.

Looking at the graphs, you can see from the "Total" line in the "Stories" graph that at the beginning, the coverage was being driven by the "deserter" charge, but on February 3rd, the "AWOL" story spiked up and has been driving the coverage since. (You can see the spike on the "Adds" graph, and see also that we're in the middle of another, extended spike.)

Once again, those graphs are available here -- the count of stories is on top, and the adds are below. I'll try to keep them updated daily for as long as it seems worthwhile.

Updates

2/13 With no strong revelations or accusations to drive it, things slowed down yesterday for the second day in a row, although the trajectory of the curve still continues to be basically upward.

Another note on methodology: for whatever reason, it seems to take a few of days -- 3 or 4 -- for the counts on Google News to calm down and stay put, so I've started checking back 4 days and correcting the data as I go.

2/12 I've fine-tuned the methodology again to avoid the problem of stories "falling off" (i.e. aging and no longer being in the Google News catalog). Now, instead of counting all the stories each day and subtract one day's totals from the last day's total, I'm counting each day's adds directly and then adding those numbers to get the totals. I've gone back and refined the numbers using this new method, and the new graphs are, once again, available at this address, with the count of stories on top and the number of adds on the bottom.

2/11 An extraordinary day yesterday, with a huge number of "AWOL" adds (476), all due to the White House releasing some of Bush's National Guard pay records. The "AWOL" count line went almost vertical, while "deserter" stays flat (and will start to decline in a couple of days as stories start to age off of the Google News database).

2/10 The big jump up from 2/6 through 2/8 was just what I said, a spike, almost certainly related to Bush's appearance on "Meet the Press" and the AWOL subject coming up. On 2/9, adds returned to normal levels, and the slope of the "Total Stories" line flattened out somewhat, although it's still going up.