Song then seen ≠ Seen then song

If an animal is making noise, are you more likely to count it? Are you more likely to misidentify an animal if you only hear it? Could those animals be most noisy at some times of the day, or year, or in some places? If the answer is “yes” to any of these questions, you should separate your counts by song and seen.

Heard versus seen only

It is standard practise in many bird counting methods to separate seen from heard. For example, arguably the most widely used bird counting method in New Zealand is the five minute bird count method. This method, promoted by the NZ government’s Department of Conservation, asks you to separate the birds heard, whether seen or not, from those only heard:

“Birds that are first heard should be entered under H (even if they are later seen), birds that are first seen should be entered under S. Adding H and S should give the total number of birds observed.”

This allows you, or researchers using your counts, to compare heard versus seen only. This can reveal the daily and seasonal changes in bird song. In large datasets it can also show if those rhythms are shifting over time or among habitats.

Importantly, it can also allow researchers to assess whether your total counts were higher when more of the birds were singing. That would suggest that your “detection probability” was higher at times when more birds were singing. That would need to be taken into account to accurately extract trends and patterns in bird numbers from your counts. This is all useful.

Seen versus heard only, and seen first versus heard first

In my wild counts I go one step further. If I see an animal singing, I record separately whether I heard it first or saw it first. Doing this lets researchers also compare seen versus heard only and seen first versus heard first. These comparisons reveal more from my counts.

Separating seen and song counts is easy to do with my shorthand. For example, here are the options for a bellbird singing (vocalising, or v in WildCounts shorthand).

Bellb 1s (one bellbird seen only)
Bellb 1vs (one bellbird heard singing then seen)
Bellb 1sv (one bellbird seen then heard singing)
Bellb 1v (one bellbird heard singing but not seen)

Comparing seen versus heard only is useful because many birds are quite tricky to ID from song alone, especially for beginners, but they’re straightforward to ID when seen. Separating “seen” from “heard only” counts allows researchers to more accurately model observer identification error rates and so calculate more robust trends.

I had this in mind when I started counting birds. I could confidently identify New Zealand’s terrestrial birds by sight, but I knew I was going to get better over time at identifying them by their calls and songs. I wanted to be able to look at trends in the birds I saw and be confident those trends weren’t biased by my improving song ID skills. I therefore needed to separate seen versus heard only.

Comparing seen first versus heard first makes it easier to separate daily and seasonal cycles of singing from trends in numbers. That’s because it provides a minimum estimate of the birds that would have been seen regardless of their song. The “seen then song” birds can be included with the “seen only” birds when assessing whether total counts were higher at times when more birds were singing.

Using song and seen to better interpret index counts

Counting song and seen separately helps a lot when calculating trends and patterns from “index counts”. A regular five-minute bird count is a type of index count, meaning that you’re not measuring the real density of birds. You don’t expect to be able to see or hear every bird during a five-minute bird count. Instead you’re recording some unknown proportion of all the birds at site, those that are visible and/or singing. Your count is said to be an index of abundance, it’s not an estimate of the true abundance.

The assumption (hope) is that if your index goes up, that means that bird abundance has gone up. Similarly, if your index is higher in one habitat than another, the assumption is that birds are really more abundant in one habitat than the other. These assumptions only work if you’re always counting the same proportion of all the birds present at each site. Index counts go off the rails when birds are easier to see in some places, or at some times, than others.

Keeping seen and song counts separate is helpful here. Consider the situation where birds are easier to see in one habitat than another, but the density of birds in the two habitats is the same. In this case our index of seen birds should be quite different between habitats but our index of bird song should be the same. We can use this to make the right conclusion: birds are equally common in both habitats.

The same thing can happen over time. For example, as a forest restoration site grows over the years, the understory may open up so you can gradually see further. This would make birds easier to see. Your index of seen birds would therefore increase over the years even if there had been no change in real bird densities. However, if real bird density didn’t change, your index of bird song should not have changed. Counting song and seen separately lets you figure this out.

Songs and calls

Many birds and other animals make different kinds of sounds for different reasons. For example, there are songs to defend territory or attract mates, alarm calls when a predator is near, and contact calls to keep a group together. When and where animals are most likely to make these different sounds is likely to vary. Territorial songs might be most common in the spring and summer during breeding. Alarm calls will likely be most common in habitats with more predators.

This means that it’s useful to record the different types of sound separately, if you know your animals well enough to recognise their different songs and calls. For most species, nobody will know what the different types of songs and calls are used for. The behaviour of most animals hasn’t been studied that closely. Separately counting the different types of vocalisations can help to uncover the meaning of these songs and calls, while also making it possible to extract more robust trends and patterns from your counts.

In my shorthand, I separate songs (v for vocalisation) from calls (c) and note the call types in the brackets. For example, here are the sounds I separate for NZ bellbirds.

Bellb 1vxm
Bellb 1vxf
Bellb 1vj
Bellb 1c (alarm)
Bellb 1c (toot)
Bellb 1c (ack)

With practise, I’ve come to recognise songs of male (xm), female (xf), and juvenile (j) bellbirds and I consistently note three types of call (alarm, toot, and ack). When I can’t clearly assign a song to sex or age, I just use v alone (Bellb 1v). Likewise, if I’m unsure of a call type, I just use c alone. If you’re don’t know much about the sounds your animal makes, just c works fine.

Let’s wrap things up

As you’ve now seen (and heard if you were reading aloud), there are many good reasons to count seen and heard animals separately. Furthermore, there are big advantages to not just separating heard versus seen only but also counting song then seen separately from seen then song. This is easy to count out in the wild, especially if you’re counting with wildcounts shorthand.

This is not just for the birds. If you’re monitoring any wild animal that makes sounds, it is useful to separate those you just see from those you just hear, those you hear then see, and those you see then hear. It’s something that’s harder to say than to do.