how about size of social unit?  You could also give percentage frequencies of the sizes of social groups, 1, 2, 3, 4 etc.  (Personally, I think this reviewer is being a pain, but that is besides the point).

You could also give the average number of social allies, 0 for loners, 1 for pairs, 2 for individuals in trios etc.

loosely associated "satellite animals" like here:

Satellite Male Groups in Horseshoe Crabs, Limulus polyphemus

H. Jane Brockmann†,*

Article first published online: 26 APR 2010

DOI: 10.1111/j.1439-0310.1996.tb01099.x

Ethology; Volume 102, Issue 1, pages 1–21, January-December 1996

In many years of observing rhesus macaques in the Himalaya, we found that the solitary males were usually once members of the family but have been thrown out as competitors to the alpha male. They tend to inhabit the same home range as their original family, occasionally fighting with the alpha male for possession of the family, but generally keeping to the distant sidelines.

I do not have any idea of what prevails among howler monkeys, but if the loners were once part of the family and inhabit roughly the same range, then they are certainly part of the group and can, as Andreas mentions, be referred to as 'satellite members' of the family or group.

We noticed that when another group passed through the territory of the resident family, the loners also came to the aid of the family to make a show of strength, so they are not as disassociated as one might be led to believe.

"Party size" or include a definition of group as 1 or more individuals

I am a bit confused about what you want to represent here. As if you are trying to calculate average size of a troop, then  floaters should not be included. As they were or can be a part of a troop, but at a given point of time they are "pushed out" or themselves move out of the group. which will be consistent in someway to the population across the landscape. If you include these outliers in estimating the pack size, I feel it will be bringing noise to your data. If you are taking a census of  howler monkey population then definitely they should be included, but for size of troop they should not, I think.

But if you do want to mention the number of floaters which are found together at a given time, you can mention it as an average only. like there must be a number of these floaters , a threshold above which you can call it a group.

The 'satellite" best explains the situation as alluded to by Peter. Observations among social animals like kobus kob indicate that they can expel a male member but he is still associated with the original group and they share the same home range. Such animals are part of the data. Do not confuse us by introducing new terms to bring 'noise' in your equation.

Taking account of group sizes of one can be important when calculating density of animals among other things so the widely used Distance Sampling approach (http://www.ruwpa.st-and.ac.uk/distance.book/) uses the term "cluster" for groups of any size including one.

Hei Sarah, 

I think I would also go with cluster, as Ross mentions. It is an accepted term in the distance sampling literature as well, where a cluster can also have a single animal. 

Best wishes,

Navi

Dear Sarah,

I had similar problems in the past with some reviewers. I do not think there is a simple solution. Depending on the traditions of the disciplines the researchers are identified with, some may even argue that 2 individuals are a dyad and not necessarily a group. Social units or clusters are undoubtedly alternative good constructs concerning your own perspective that I also share. I am afraid that your reviewer will not be pleased anyway. Maybe you have to use the distinction between grouped and ungrouped individuals. You could provide first both proportions of individuals and subsequently calculate the mean group size just for the grouped ones.

Best regards,

António

Dear all, many thanks for all your helpful answers!

I agree that the lone howler monkeys may be 'satellite' individuals associated with particular groups.  However, without a much longer term study of the social relationships between all the individuals we counted, I don't think we can use the term 'satellite' in the paper, as using the data we have, we can't demonstrate that the lone animals are associated with specific groups. 

I think I will use 'social unit' or 'cluster'. I think perhaps 'social unit' may be more acceptable to the reviewers, but perhaps 'cluster' may be more appropriate if it is more widely used and suggests a geographic association, without assuming specific social relationships. Referring to 'social units' sort of suggests that lone individuals never socialize with other individuals, which we cannot prove.

Dear Sarah,

Importantly, distinguish the descriptive question of what the sizes of groups are, from the functional question of *why* the groups are the sizes you observe, and - particularly for your case - why some individuals are alone. And, unless you are confident about function, use a neutral term, for the job of description. This is good science; you need *evidence* to use a term that assumes (or even implies) a function. (The same logic holds for naming behaviour patterns.) So, for your descriptive term I suggest simply 'group' or something similarly neutral with regard to function. And be explicit about how you are using terms (descriptively or functionally).

Turning to your question of whether to include solitary individuals in a measure of 'mean group size' it depends on your question. Since your question is 'how many howler monkeys are you likely to find together' I agree that you should include the singles. A simple extension of your question (which I'm sure you will be asking ) is 'what is the frequency distribution of the numbers of animals found together' (and you should have a robust definition of 'together')? Again, you should include singles. And when you look at the the shape of the frequency distribution you will get some weak clue as to function; if the frequency of singles is what you would expect from the fit of a simple curve (e.g. normal, poisson) to the rest of the distribution then perhaps the simplest explanation for the distribution you find (and, in particular, for the frequency of singles) is that these are randomly forming groups (probably the same as what Joel Cohen called 'casual groups'; see his book 'Casual Groups of Monkeys and Men'). In this case, crudely put, singles are nothing special. But the truth may be far more interesting than this . . . 

I suggest using "party size." It has been used widely for studies of fission-fusion species, especially in chimpanzees.  "Social unit" implies that the group has a functional purpose and may be a permanent grouping, which may of may not be the case for howler monkeys.  "Clusters" is not commonly used in the socioecological literature for observed group sizes.  Rather, "cluster" tends to be used more when you are pooling information from multiple observations of groups to determine which animals associate together in higher-order communities (such as elephants and giraffes).  

I also agree that you cannot exclude the 1s from any analysis of group size distributions, nor can you conclude that these animals are truly satellites without more data.  Perhaps you can keep the 1s in for most of your analysis, but to satisfy the reviewer,  add a sentence saying something like "50% of observations were of a solitary animals, and the mean group size excluding solitaries was....."  Personally, I have never had an issue of using group size when groups sometimes only had one individual.

Group composition table may help you to show how many individual monkey you sighted. you may use the name 'cluster size' and /or 'mean crowding size' these terms are widely used.  

Group size certainly includes number of individuals in troop living cohesive manner. The solitary individuals may not be included in a group but they may be counted and considered when population size is described.

If you are just worried about the statistical representation, a group of 1 is still valid if any monkeys are ever found alone. You are talking about a sampling unit and 1 is just part of the distribution. It sounds like the reviewer is hung up on semantics rather than context....

If you want an example with social animals that often appear alone, I refer to 1 as part of the range of group sizes here:

Ransom, J. 2012. Detection probability in aerial surveys of feral horses. Journal of Wildlife Management 76: 299-307.

You should be able to get the PDF on my RS page, but if not shoot me an email.

good luck-

Jason