I think older obs tend to have more misidentifications because of the CV as well as an accumulation of experts and knowledge passed around the community, but only if they haven’t gotten attention in the years since. My impression is that most people usually start with recent and work backwards. Varying differences in identifier time vs observations, leave pockets of NeedsID and misidentifications that get missed until random chance or thorough review gets to them.
There’s some taxa where I’ve set the identification interface filter end date to earlier than today, sorted by date added, descending and worked from the oldest page to newest page to try to miss as little as possible. Sometimes it looks very accurate and well-looked at from the earliest until a few years ago or now. Sometimes it’s spotty with a few that looked missed because of the way thing can get skipped when obs are added or change status. Sometimes there’s sections of months or more that apparently didn’t get much attention. Sometimes its just a couple of the oldest pages of observations with clear inaccuracies and the rest appears well-curated.
Tested this formally? No, but I have done exhaustive identifying of all observations of a number of well-observed lizard species (I think I have over 1000 IDs on over 25 species of lizards). Here’s what I’ve noticed.
Newly posted records seems to be misidentified more than older observations probably because there are more recent observations and more eyes have been on the older observations, as you suggested. However, I have found a number of observations with multiple agreeing identifications (three, four, five, I think I even found an observation with seven wrong IDs once) that were (in my opinion) incorrect. Mainly it was when species were visually similar to another closely related, but sympatric species in the area. For example, hatchling and juvenile iguanids can be notoriously hard to identify based on color or pattern, but their tails and throats are reasonably recognizable making them ‘easy’ to identify with practice. So older observations aren’t free from errors, but they probably are less error-prone in some situations (if the CV became aware of that species early on, if that species has a champion, if it is ‘easy’ to recognize, etc).
The other thing to consider you didn’t mention is more specifically about the community of identifiers in that region. Most users identify regionally at best meaning a single area usually has the same identifiers that slowly get replaced geographically. So, if that community isn’t as strong at catching that initial errors, they get buried in four observations (that number of IDs showed again and again it seems) all of which are incorrect. So, the age of the observation doesn’t seem to allow for easy correction unless someone is a “complete-ist”, like me.
So, I think older observations have way fewer errors numerically (or even proportionally) merely because more time has elapsed for correcting ID issues and solid identification communities have developed in many geographic and taxonomic areas. For example, the community of lizard identifiers in the U.S. was strong before I even found it.
Good observations. I’ve noticed similar trends with spiders–older observations are more likely to be identified correctly, but when they’re identified incorrectly, it’s often a pile on of 3-7 identifiers confidently misidentifying. Sometimes it’s clear they did some research (such as by posting links to another site that shows a different, but similar-looking, spider). Most of the time, it’s clear they were just agreeing randomly (total of 17 IDs, each of a totally different taxa). I think when there were fewer observations in total, the impact of newbees playing around with the “agree” button had a larger impact. Now, that still happens today, but most of the misidentifications I see are the observer following the CV, rather than 7 inexperienced users hitting agree on a guess.
If you do some identifying, I definitely recommend using the Random option for Sort in the filters, which will bring up older observations for scrutiny.
I don’t use the Identify tab to identify, but rather the Explore tab. The random sort option doesn’t appear possible through the Explore portal’s filters.
Perhaps, but the lack of geographic up front on the Identify tab is problematic in that I usually identify in small geographic areas at a time. Not that “place” can’t be added to the search, but once in an individual observation the map becomes the problem. In some cases (and this is only getting worse), the number of orange observation squares nearly completely obscures the map and the map overlays aren’t sticky meaning to get rid of that effect one must remove a taxon’s past observations each and every time in some (many) regions. Here’s an example. I’ll stick to the Explore tab.
There is a way to do this in Identify, but admittedly it is a multistep process. First, select the portion of the map that you want to work on; then switch to grid view; then select “identify” at the bottom of the “filters” menu.
