The Number of Identified Specimens or Number of Individual Specimens (NISP), is used in archaeology and paleontology when counting bones from a site. NISP counts each bone and fragment as one unit.
NISP can often be an overestimate of the actual number of individuals at the site, especially when preservation is good but bones are highly fragmented. Multiple fragments of the same bone lead to it being counted multiple times. However, too much fragmentation can lead to an inability to identify a bone as a particular type in the first place. At many sites with poor preservation the total NISP severely underestimates the total possible number of individuals that contributed to the assemblage. Some paleoindian sites, which often exhibit poor organic preservation, are a case in point. However, this and some other problems can be corrected analytically.
A more serious problem is the one of interdependence. There is generally no reliable way to determine if individually counted specimens originated from the same organism on the ancient landscape and are therefore interdependent. Since dependence is generally possible, NISP consistently overestimates, and to some unmeasurable extent, the actual number of individual animals represented in a recovered assemblage.
An alternative estimate to NISP, often done in concert, is minimum number of individuals (MNI). Both are influenced by fragmentation and limited preservation, but in different ways. NISP tends to overestimate the number of individuals under moderate fragmentation, but the overestimate lessens as fragmentation increases due to the inability to classify the bones. MNI tends to underestimate the actual number under medium fragmentation, and even more severely when bones are highly fragmented. Under hypothetically perfect preservation and no fragmentation, these estimates should be the same. MNI also suffers from the aggregation problem, in which different aggregations will generate at least two values, a MNI minimum and maximum, which are generally empirically indistinguishable. Both NISP and MNI are likely only ordinals scale measurements, which means at best they can only give an ordered series of taxonomic abundance, i.e. "Taxon A is more numerous than Taxon B."
NISP should not be used when calculating a sample size for inferential statistics, because it will inflate the statistical significance. Thus in these situations MNI should be used instead.