The NCI are a series of four islands located off the coast of California in the Santa Barbara Channel. From west to east, they are San Miguel, Santa Rosa, Santa Cruz, and Anacapa Islands (Figure 1). Sea level rise, beginning after the Last Glacial Maximum and continuing through the initial periods of human occupation of the islands, led to an ~85% reduction in land mass from the date of the first evidence of people on the NCI around 13,000 until ~6,000 cal BP, when marine transgression slowed (Kennett et al., 2008; Reeder-Myers et al., 2015). At the time of earliest occupation, lower sea levels would have exposed land connecting the four NCI into a single larger island, Santarosae (S^hamalaḿa; Rick et al., 2023). The NCI and the surrounding marine ecosystem have provided islanders with a great abundance and diversity of subsistence resources throughout their history. This is especially true along what is today the northwest coast of Santa Rosa Island (SRI), which contains large, well-watered drainages that are especially resilient to periods of drought, as well as extensive rocky intertidal coastlines where shellfish can be collected en masse. It is for this reason that previous models have indicated that the habitats there would have been high-ranked, and therefore preferred settlement locations, through time (Kennett et al., 2009; Winterhalder et al., 2010; Jazwa et al., 2016a,b). This is likely among the reasons that the oldest dated site on the NCI, Arlington Springs (CA-SRI-173), is located there (Johnson et al., 2002).

However, the earliest sites on the NCI, those dating to the Paleocoastal Period (13,000–8,200 cal BP; terminal Pleistocene and Early Holocene), are not consistent with what Kennett (2005) and others (e.g., Jazwa, 2015) have described as permanent settlement sites, namely those containing dense, accumulated midden reflecting year-round occupation or consistent reoccupation. Instead, most early sites on the NCI that still survive tend to be more ephemeral occupations located along the edges of elevated coastal terraces (e.g., Erlandson et al., 2011, 2020; Rick et al., 2013), although some prominent sites containing characteristic bifacially flaked artifacts and points (crescents, Channel Islands Barbed, and Amol) have been identified closer to the coast (Erlandson et al., 2011; Erlandson, 2013). Most of the coastal regions that would likely have been targeted for early settlement, however, are now underwater, submerged by rising sea levels, making developing habitat suitability models for the terminal Pleistocene and early Holocene difficult (Kennett et al., 2008; Reeder-Myers et al., 2015). One recently investigated site at the mouth of Trancion Canyon, CA-SRI-347, contains midden deposits consistent with permanent occupation dating between ~11,000 and 8,500 years ago (unpublished radiocarbon dates obtained by Jazwa). This site is unique because it is located adjacent to an area of steep bathymetry where the shorelines have likely not moved far, providing further evidence that additional early permanent settlement sites likely exist in locations that have been submerged. Nevertheless, according to the IFD, habitats that were occupied earliest should continue to be occupied as settlement expands, so even if the earliest phases of occupation have been lost to rising sea levels, the earliest available data on settlement locations should be indicative of preferences in population distribution.

For this reason, the earliest clear evidence for sustained occupation of the coast of SRI dates to the beginning of the middle Holocene (8,200–4,200 cal BP). The model developed by Jazwa et al. (2016b: Figure 9) includes speculation as to what settlement distribution may have looked like before this time, but it was not possible to directly test model predictions for the very earliest settlement on the island. Other than CA-SRI-347, among the earliest dated permanent coastal residential sites on SRI are CA-SRI-187 at the mouth of Old Ranch and Old Ranch House Canyons (8,050–7,040 cal BP), CA-SRI-138 at the mouth of La Jolla Vieja Canyon (7,545–7,415 cal BP), CA-SRI-3 at the mouth of Tecolote Canyon (7,890–7,520 cal BP), and CA-SRI-4 at the mouth of Arlington Canyon (7,470–6,920 cal BP; Kennett, 1998, 2005; Winterhalder et al., 2010; Jazwa et al., 2013, 2016b; Jazwa, 2015). While marine transgression continued after the earliest occupation of these sites, it was limited enough that they would still have been adjacent to or within several 100 meters of the coast at any period for which there is evidence of occupation and could therefore be considered coastal settlement sites. Furthermore, with the exception of CA-SRI-3 and CA-SRI-4, which are associated with adjacent drainages, these sites are located relatively far apart from each other (7.5–10 km from each one to the next closest) and adjacent to different environmental zones, including drainages expected to be high-, middle-, and low-ranked for settlement (Jazwa et al., 2016b). Still, compared to some locations on the NCI, especially parts of northern Santa Cruz Island, all habitats on SRI are relatively high-ranked and able to support small populations (Kennett et al., 2009; Winterhalder et al., 2010).

During the Middle Holocene, populations increased and more permanent villages were established on SRI. These primary villages were mostly located on productive coastal habitats with access to fresh water from springs or streams (Kennett, 2005; Winterhalder et al., 2010; Jazwa et al., 2016a). In some cases, these water sources were seasonal. While the Early Holocene was generally a wet period in California and the Great Basin (Benson et al., 2001, 2002, 2013), climate records from throughout the region suggest that the middle Holocene was overall a relatively dry interval (e.g., LaMarche, 1973, 1974; Lindström, 1990; Thompson, 1992; Quade et al., 1998; Hughes and Graumlich, 2000; Benson et al., 2002; Kennett, 2005; Kennett et al., 2007). During drier periods, the large, resilient drainages of the northwest coast of SRI would have been especially valued locations for human occupation (Jazwa et al., 2016a). As the island population increased during the Middle Holocene, settlement focused on that part of the island, with people becoming more sedentary to take advantage of the northwest coast to access fresh water and its highly productive coastline. The archaeological record is consistent with this prediction, with evidence for Middle Holocene settlement at the mouth of Tecolote, Arlington, Soledad, Dry, Verde, Cow, and Lobos Canyons along the north coast (Kennett, 1998; Winterhalder et al., 2010; Jazwa et al., 2016b). This increased focus on north coast settlement is associated with a greater diversity of site types, with patterns of interior residential bases accompanying coastal settlements and special purpose shellfish harvesting and shucking sites (Kennett and Clifford, 2004; Kennett, 2005; Jazwa et al., 2015; Jazwa and Zoellner, 2024). Movement to the interior during the summer and fall may have also been related to fog water collection or terrestrial plant harvesting (Jazwa et al., 2015, 2016a).

During the Middle Holocene, red abalone middens come to characterize many sites across the NCI (Glassow, 2013). Red abalone (Haliotis rufescens) is a large marine gastropod that is among the highest return food species on the NCI (Braje et al., 2007). In southern California, they tend to live in subtidal environments, but may live in shallower, intertidal regions during periods of cooler water (e.g., Raab, 1992; Glassow et al., 2012). There is a debate over the reason for the appearance of these sites during this period. Glassow (1993; Glassow, 2013; Glassow et al., 2012) has argued for an environmental explanation, in which red abalones were more readily available because sea surface temperature (SST) was colder at this time (Kennett and Kennett, 2000; Kennett et al., 2007). Conversely, Braje et al. (2009; Erlandson et al., 2004, 2005, 2008) argue that red abalones were more common during the Middle Holocene because they were released from predation pressure from otters, who were increasingly hunted by humans. Both may have influenced the population of accessible abalone at this time, but notably, the location of red abalone middens on Santa Rosa Island do not necessarily follow the predictions of settlement rank models (Jazwa et al., 2015, 2016b; Jazwa and Zoellner, 2024). Put another way, the factors that influence where abalone are present may or may not track the variables that influence the location of permanent settlement sites. As Gamble (2017) points out, however, there is great diversity in the nature and use of the red abalone middens, including residential, special purpose, and potentially even feasting sties. Jazwa et al. (2015) found the highest concentration of red abalone in an otherwise relatively low-ranked habitat, Bee Canyon, suggesting that sites like CA-SRI-338 were used primarily for abalone processing before transportation back to permanent settlements elsewhere on the island.

From 4,200 to 1,500 cal BP, sea surface temperature became more variable (Kennett and Kennett, 2000). McFarland et al. (2025) have argued that this led to a change in the relative proportions of marine invertebrate taxa on the NCI, especially an increase in the average proportion of urchins compared to other kelp-dependent invertebrates. Furthermore, people living in the Santa Barbara Channel region developed the single-piece fishhook by around 2,500 cal BP (Rick et al., 2002), immediately preceding a dramatic increase in fish in the archaeological record (Glassow, 1993; Kennett, 2005; Jazwa et al., 2016b, 2020). Both of these changes are consistent with changes in the faunal record from the NCI. The general pattern of climatic amelioration during the Late Holocene (Kennett and Kennett, 2000; Kennett et al., 2007) coincided with increases in population density, the number of settlement sites, and the range of habitats occupied on the islands (Kennett, 2005; Kennett et al., 2009; Winterhalder et al., 2010; Jazwa et al., 2016b, 2019). By the start of the Medieval Climatic Anomaly (MCA; ~1,150–650 cal BP), for example, settlement had expanded to the mouths of all 19 large drainages on SRI (Jazwa et al., 2019). During the MCA, a period of extreme and persistent droughts occurred in the southwestern United States and NCI settlement structure changed dramatically. Overall island population densities continued to increase, though these appear to have been concentrated in a smaller number of large coastal village sites. These village sites were often in new locations, typically across the drainage mouths from preexisting settlement sites. These sites are often the large villages named in early 19^th century mission records (Johnson, 1993; Glassow et al., 2010). The villages generally had high population densities and their political economy has been described as that of a complex chiefdom with an economy characterized by sophisticated sewn plank canoes called tomols and purple olive shell (Callianax biplicata) beads produced by craft specialists (e.g., Arnold, 1992, 2001; Kennett, 2005).

The NCI have proven to be a productive test area for applying and refining applications of IFD/IDD models. Anthropological applications of the IFD first appear in the literature in Kennett's (2005) book in theoretical terms, as speculation as to how they could be applied to settlement on the NCI. After a similarly theoretical application for the settlement of Oceania (Kennett et al., 2006; Allen and O'Connell, 2008; revisited in O'Connell, 2020), the first formal test of the predictions of the model using ecological and chronological data was in Kennett et al. (2009). Winterhalder et al. (2010) followed that up by using a Bayesian statistical approach applying the chronology of initial settlement for drainages with dated sites to assess whether the suitability calculations by Kennett et al. (2009) appropriately weighted natural resource availability for the different habitats. Since that time, the IFD/IDD have been applied to a great diversity of contexts around the world, including elsewhere in North America, the Caribbean, Mesoamerica, Greece, and northwest Africa, among other places (e.g., Giovas and Fitzpatrick, 2014; Prufer et al., 2017; Jazwa and Jazwa, 2017, 2022; Harvey, 2018, 2019; Yaworsky and Codding, 2018; Jazwa and Collins-Elliott, 2021; Collins-Elliott and Jazwa, 2022; Hanna and Giovas, 2022; Plekhov and Levine, 2022; Weitzel and Codding, 2022). However, despite the initial studies on the NCI demonstrating that settlement distributions are largely consistent with IFD predictions, further studies have used the archaeological record there to refine nuances of the model. The NCI are particularly suitable for this purpose because they are bounded by water and therefore we can assess a complete set of potential habitats, settlement locations can generally be assumed to be individual points at the mouths of prominent drainages based on our knowledge of settlement on the coast, subsistence resources are limited in range (but not quantity) compared to mainland contexts, and there is a long history of archaeological research with good information about settlement history of most of the settlement locations. Since the initial applications, further research has been conducted to better understand the effects of population-independent environmental change (Jazwa et al., 2013), apply the zooarchaeological record as a proxy for population density (Jazwa, 2015; Jazwa et al., 2016b), and assess the transition from the IFD to IDD under periods of resource stress and population restructuring (Jazwa et al., 2019).

In this study, we use the NCI archaeological record to better understand how the IFD can be used to model settlement at very low population densities. A fundamental prediction of the IFD is that initial settlement will be in those habitats ranked highest (e.g., Jazwa et al., 2016b). As population density increases, settlement will expand to progressively lower ranked habitats, with successively lower-ranked habitats settled and “filling in” with human population until the ratio of people to resources in each habitat makes per capita returns in each habitat effectively equal. However, Jazwa et al. (2016b) found that the first prediction regarding initial settlement was only partially supported on the NCI. Some lower-ranked habitats, namely those along the south coast of SRI, were first settled much earlier than predicted by the model. In fact, settlement distribution best matched the IFD only beginning during the Middle Holocene when intensive settlement began to focus on the high-ranked, northwest coast of SRI. Jazwa et al. (2016b) attempted to explain this in two ways. First, they argued that the earliest settlements on the island were not necessarily the earliest permanent settlements; those were likely submerged by Pleistocene and Early Holocene sea level transgressions. Jazwa has obtained radiocarbon dates reflecting earlier evidence for permanent settlement at the mouth of Trancion Canyon, the third highest ranked drainage overall, supporting this idea. More importantly, however, Jazwa (2015) speculated that at low population densities, settlement decisions may be governed by different considerations than those used in previous iterations of IFD modeling. Population density was low enough that individuals could maximize their return at a wide variety of habitats. Unlike later in time, there was no need for people to position themselves in habitats that would maximize overall resource return. Rather, they could have chosen from a large number of habitats that all could have provided more subsistence resources than what population at that size can exploit. In the case of the NCI, when considering all potential habitats for Santa Cruz, San Miguel, and SRI, there are no SRI habitats among the lowest ranked 1/3 of all drainages across the islands. Low-ranked drainages that may be resource limiting, even at low population densities, are confined to Santa Cruz and San Miguel Islands (Winterhalder et al., 2010).

Therefore, it may be the case that settlement decisions at low population densities may not be guided by efforts to maximize resource return. Instead, in those cases, the components of the model associated with calculating habitat suitability may only be relevant to the extent that resources meet a certain threshold. In those cases, settlers may choose between any of those habitats based on a different set of criteria that are otherwise not considered of high importance once population density increases to the point that it is beneficial to maximize resource return. This decision may be based on less quantifiable considerations as simple as propensity for warm, sunny days and/or less frequent storms. Alternatively, a common assumption of the IFD is one of complete knowledge of the region under consideration, including sufficient information to accurately appraise and compare relative suitabilities for all habitats (Winterhalder et al., 2010: 471; Assumption f). This may not necessarily be the case, especially in a large geographic area, and a settling population may be willing to settle down in one of the first habitats that they encounter that are good enough to provide the subsistence resources they need in quantities that are more than they can consume, even if other, nearby habitats allow for greater resource maximization (Haas and Kuhn, 2019).

Shell midden deposits included in this study were excavated in two phases. In 1996 and 1999, Douglas Kennett and Don Morris excavated four 25 × 25 cm units from CA-SRI-130, two 25 × 25 cm units and one 50 × 50 cm unit (Unit 3) from CA-SRI-131, and four 25 × 25 cm units from CA-SRI-135¹ (Figure 1). In August 2019, we excavated one 25 × 25 cm unit from CA-SRI-139 from an eroding cliff-face with Richard-Patrick Cromwell and Gilbert Unzueta. We excavated in arbitrary 10 cm levels, separating materials from clearly defined strata and collecting radiocarbon samples in the field from the wall profile after excavation. We also included previously published comparison data for one 25 × 25 cm unit each from CA-SRI-131 and−488, and two 25 × 25 cm units from CA-SRI-138 (Jazwa, 2015; Jazwa et al., 2016b). We excavated using the same methods as CA-SRI-139. Kennett and Morris also obtained radiocarbon dates that we include in this analysis from further excavation units at CA-SRI-147 (Braje et al., 2007), CA-SRI-62, and CA-SRI-432, and field sampling from eroding escarpments at CA-SRI-419 and CA-SRI-462. We collected samples from eroding escarpments at CA-SRI-63, CA-SRI-144, and CA-SRI-145.

We include 38 radiocarbon dates from the south coast of Santa Rosa Island in this analysis, 12 of which are new dates obtained for this study. Of those dates, 24 are from sites for which we are including excavation data and the remaining 14 are from nearby sites. For the 12 dates new to this study, we cleaned the samples, subjected them to a 50% hydrochloric acid etch, hydrolyzed them to generate carbon dioxide at the UNR Human Paleoecology and Archaeometry Laboratory, and submitted them for measurement for radiocarbon at the Pennsylvania State University AMS Laboratory. Original references for the 26 previously published dates are provided in Table 1.

We calibrated all the available dates for the south coast of Santa Rosa Island in OxCal 4.4 (Bronk Ramsey, 2009) using the Marine20 calibration curve (Heaton et al., 2020). We used a variable ΔR value for the Santa Barbara Channel region for samples with dates of < 2,500 radiocarbon years before present (Hendy et al., 2013) and an estimated ΔR value of 124 ± 30 for older samples obtained from six historical, pre-bomb Callianax biplicata shells. All ΔR values have been corrected for the Marine 20 calibration curve (Sean Hixon, Brendan Culleton, and Douglas Kennett, personal communication, 2021).

For our analysis, we assess settlement expansion following a modified version of Arnold's (1992) chronological framework to better distinguish important periods of change during the Late Holocene: Early Period (7450–2550 cal BP), Middle Period (2550–800 cal BP), Middle to Late Period Transition (MLT; 800–650 cal BP), and Late Period (650–168 cal BP). Notably, the droughts of the MCA occurred at the end of the Middle Period through the MLT (i.e., Late Middle/MLT). The Early Period was not uniform, so we distinguish the periods before and after the Middle and Late Holocene boundary at 4200 cal BP for the purposes of this study.

After excavation, we size sorted the midden through nested 1/2”, 1/4”, and 1/8” screens. Materials from sites excavated in the 1990s were preliminarily sorted at University of California, Santa Barbara and California State University, Long Beach by trained undergraduate students under Kennett's (1998; 2005) supervision. We similarly size-separated the material from CA-SRI-139 prior to sorting the midden to taxa. We sorted all the 1/2” material to artifact type or taxa, in most cases assigning shellfish to the species level. For the purposes of this analysis, we do not more precisely identify fish and sea mammal bone. For the smaller material, we similarly sorted subsamples, 100 g for 1/4” and 15 g for 1/8”, excluding non-cultural rock. For residual bulk midden samples, we sorted for diagnostic artifacts, bone, charcoal, ochre, Mytilus californianus hinges, and Callianax biplicata shell. We used the total weight of bulk midden and proportions of the subsamples to estimate overall taxonomic contributions for each level. We obtained the data published in Jazwa (2015) and Jazwa et al. (2016b) following the same protocols. We combined data from stratigraphic levels of the same units into associated contexts, with a single excavation unit comprising either one or two contexts depending on radiocarbon date sequences and similarities in faunal constituents between levels. For comparison, we normalized faunal weights in each context to one cubic meter. We calculated the Shannon-Weaver diversity index (Shannon and Weaver, 1949), evenness, and taxonomic richness for shellfish in each context using these compiled shell weights.

We use meat weight calculations summarized by Rick (2004; see also Colten, 2001; Jazwa et al., 2019) to estimate the relative contribution of shellfish, fish, sea mammal, and bird to the diet for each of these components. Meat weight calculations have important limitations, especially for sea mammals that are likely to be field processed. Not all elements have the same meat to bone ratio, which might bias the contribution of sea mammal to the diet (Mason et al., 1998; Gifford-Gonzalez and Hildebrandt, 2012). Nevertheless, the great difference between the ratio of bone/shell to meat for fish, sea mammal, shell, and bird makes it difficult to compare them based on faunal weight, so we cautiously apply meat weights for this comparison. Furthermore, the fact that all the sites considered in this analysis are located on the coast likely reduces the extent of field processing.

Finally, we estimated mussel size for those incomplete shell fragments with the umbo present using the equation provided in Singh and McKechnie (2015)². For complete shells larger than 20 mm long or fragments of shells estimated to be >20 mm, we obtained the average shell length by context. To avoid counting individual mussels twice, we only included values from those shells of the side with a higher count from each stratigraphic level (i.e., left or right valves). We also calculated the relative proportion of mussels within each assemblage that were < 20 mm long.

Based on the new and previously published radiocarbon dates, it is clear that permanent settlements were established at the mouths of both La Jolla Vieja and Wreck Canyons earlier than would be predicted by the IFD. All dates are given as 95.4% calibrated ranges. Early occupation at CA-SRI-138, Unit 1 dates to 7555–7270 cal BP (Table 1; Figure 2). Among permanent settlement sites, only CA-SRI-187 at the mouth of Old Ranch and Old Ranch House Canyons (Jazwa et al., 2013), CA-SRI-3 at the mouth of Tecolote Canyon (Winterhalder et al., 2010), and CA-SRI-347 at the mouth of Trancion Canyon (Jazwa, unpublished radiocarbon dates) date earlier. All three of those drainages are much higher ranked than La Jolla Vieja. There is a reversal in the radiocarbon dates associated with the terminal occupation of CA-SRI-138, Unit 1, with the sample from 49 cmbs dating to 4895–4540 cal BP, but samples from 9.5 and 10 cmbs date to 5910–5595 and 6110–5740 cal BP, respectively. This suggests some level of mixing in the upper component of this unit, perhaps resulting from erosion from an upslope, adjacent area of the site. However, settlement at Wreck Canyon is also early. A radiocarbon date from CA-SRI-462 places occupation at 7090–6600 cal BP, and one from CA-SRI-139 dates to 6625–6285 cal BP. This date is not from the excavation unit at the site. The bulk of occupation from that unit was from late in the Middle Holocene, dating from 5235–4890 to 5190–4855 cal BP. A lower density of deposits indicates persistent occupation at the site as late as 3185–2830 cal BP.

Of the three drainages considered in this study, La Jolla Vieja exhibits the greatest evidence of persistent occupation through the Middle and Late Holocene. CA-SRI-147, the radiocarbon and faunal record of which was reported by Braje et al. (2007), was occupied as early as 7410–6950 cal BP and persisted until 455–140 cal BP. This site is a rockshelter with at least three meters of dense cultural deposits, but it is ~1 km up the drainage, meaning it is not a coastal settlement. Nevertheless, it was clearly a focus of occupation for a long time. CA-SRI-130 and CA-SRI-131 are located on either side of the drainage mouth, and both have evidence for occupation dating to the beginning of the Late Holocene, with CA-SRI-131 occupied as early as 3670–3350 cal BP and CA-SRI-130 as early as 3435–3110 cal BP. Occupation of CA-SRI-130 ended before the start of the MCA, but it persisted at CA-SRI-131 until 715–375 cal BP. There is no evidence for occupation of Wreck Canyon after the low density of occupation at CA-SRI-139 dating to the beginning of the Late Holocene.

Though not directly associated with any of the three drainages, there is a Middle Holocene presence at Johnson's Lee to the west, with CA-SRI-145 dating as early as 4725–4320 cal BP. The other dates currently available from Johnson's Lee correspond to the Late Holocene, with the latest from CA-SRI-144 dating to 2320–2000 cal BP. However, Johnson's Lee is the likely location of the historically-named village of Nilal'uy, perhaps at CA-SRI-62, and therefore was likely occupied until the time of Spanish contact (Kennett, 2005; Glassow et al., 2010). Unfortunately, that part of the island and that site were heavily impacted by the construction, use, and razing of an Air Force Base that was in use from 1952 to 1963, before modern archaeological investigations were possible. CA-SRI-432, at Ford Point between San Augustin and Wreck Canyons, was occupied from 2040 to 1460 cal BP. Finally, CA-SRI-135 at the mouth of San Augustin Canyon, is the potential location of the historically-named village of Helewashkuy (Kennett, 2005; Glassow et al., 2010). The radiocarbon dates are consistent with this, with the earliest dating to 490–155 cal BP. Dates from the four units from the site are all within the last 500 years before present and the units vary in depth but are highly dense. It is possible other sites near the mouth of San Augustin may date earlier, but we have not yet been able to obtain earlier dates from any of them.

We present data from 19 distinct components from six different sites originally collected by Jazwa (2015) and Molter (2021). Over time, there is a general decrease in the relative contribution of shellfish to the diet along with an increase in sea mammal and especially fish (Table 2; Figure 3; Supplementary Table 1), a pattern common on the NCI (see Glassow, 1993; Colten, 2001; Jazwa et al., 2016b, 2020; Hagerty, 2024). Overall, the densest estimated meat weights are from the latest deposits, the Late Middle/MLT occupation of CA-SRI-131, Unit 2 and the Late Period village components from CA-SRI-135, along with the Middle Period component from CA-SRI-138, Unit 2. This is consistent with the late expansion of settlement to low ranked habitats discussed in Jazwa et al. (2016b; also Winterhalder et al., 2010), and the development of a limited number of high-density villages during the Late Period, with locations influenced by territorial behavior (Jazwa et al., 2019). The large amount of fish at CA-SRI-131, Unit 2 (73.2%) and at CA-SRI-135 (more than 85% of total meat weight in each of Units 1–3) is consistent with late settlement sites elsewhere on the island that relied on fish to support higher population densities (e.g., Jazwa et al., 2020).

During the earliest occupation dating to the beginning of the Early Period, however, there is nevertheless dense occupation at CA-SRI-138, Unit 1 near the mouth of La Jolla Vieja Canyon, with total estimated meat weight density of 241.7 kg/m³ in the deeper component and 172.1 kg/m³ in the upper component. These components have the 8^th and 10^th highest meat weight densities of all 19 dated components we have for the south coast of the island, meaning that meat weight density does not appear to increase in a unilinear trajectory over time. The later deposits at CA-SRI-138, Unit 2 increase in density into the later part of the Early Period and the Middle Period, suggesting an increase in population density at that location over time. However, there is a gap in occupation between the latest date of occupation at Unit 1 (4895–4540 cal BP) and the earliest date of occupation at Unit 2 (3180–2020 cal BP). This may indicate that the site was temporarily depopulated and later reoccupied in a manner consistent with settlement expansion to lower-ranked habitats over time, per the IFD.

CA-SRI-139, at the mouth of Wreck Canyon, was also occupied relatively early, with its most intensive occupation phase during the Middle Holocene. While its deposit was not as dense as the earliest occupation at La Jolla Vieja Canyon, the overall meat weight density of the denser, earlier phase at the site was 12^th overall at 158.5 kg/m³. However, both CA-SRI-139 and CA-SRI-138, Unit 1 indicate an almost complete reliance on shellfish, with very little fish or sea mammal in the record. In fact, the three Middle Holocene occupation phases associated with these sites are among the five densest in terms of shellfish meat weight density (2^nd, 4^th, and 5^th) among those considered here. Clearly, then, there is an important focus on these habitats early in time that persists, but only grows slowly through time during the Middle Period, MLT, and Late Period.

When focusing on the particular shellfish species present at the components considered here, it is clear that California mussel (Mytilus californianus) is the dominant species through time (Table 3; Figure 4). This is not unexpected as it follows the pattern commonly observed in faunal analysis of most sites on the NCI (Colten, 2001; Rick, 2007, 2011; Braje et al., 2007; Erlandson et al., 2009; Jazwa et al., 2015, 2016b, 2019; Jazwa and Zoellner, 2024). The Middle Holocene component of CA-SRI-139 has an especially high density of California mussel (251.2 kg/m³), second only to the Middle Period occupation of CA-SRI-138, Unit 2. Black turban snails (Tegula funebralis) are also especially prominent at CA-SRI-139, far more abundant than at any other site in the study area (118.2 kg/m³), consistent with some contemporaneous sites on the north coast of SRI (Jazwa et al., 2015). Purple urchin (Strongylocentrotus purpuratus) and platform mussel (Mytilisepta bifurcata) are among the other prominent species in the study area.

Diversity of shellfish taxa tells an interesting story of dietary change on the south coast of SRI (Table 4; Figure 5). Diversity is highest during the Late Period, which makes sense considering that was the time when populations were highest (greatest at CA-SRI-135, Unit 2; 1.8). However, diversity was also relatively high early in time during the Early Period. The decrease in shellfish diversity during the Middle Period does make sense in the context of an increasing focus on fish and sea mammal at the time. However, the high diversity at the earliest occupation of CA-SRI-138, Unit 1 (1.4), indicates that the population targeted a wide range of species in moderate amounts early in time. This is indicative of a substantial population, although not nearly as large as the density of people present later in time. Taxonomic richness is relatively high early, with 16 taxa present at the Middle Holocene deposits from CA-SRI-139. There is variability in richness later toward the end of the Early Period and throughout the Middle Period, despite overall lower diversity. This suggests that while even though people at some sites like CA-SRI-488 targeted a relatively high number of different taxa (19), the focus was on a relatively small number of species, as reflected by the diversity score of 0.7. During the Late Period, as represented by CA-SRI-135, there is generally a high taxonomic richness, 20 at Unit 1 and 17 at Unit 2, although it is lower at the relatively shallow Units 3 and 4.

Finally, we provide the estimated average mussel size by context and relative proportion of mussels within each assemblage that were < 20 mm long (Table 5; Figure 6). This provides a proxy for the likelihood that mussels were collected by plucking individual shells or stripping them from the rock. The latter is more likely to occur when fewer large, high-return mussels are present (Jones and Richman, 1995; Bettinger et al., 1997). In general, average mussel size is highest at the earliest occupied components, including CA-SRI-138, Unit 1 (49.9) and CA-SRI-139 (54.3). These are consistent with average sizes for components with high average mussel size elsewhere on the island (Jazwa et al., 2016b). Overall, mussel size decreases later in the Early Period and into the Middle Period, with the highest proportion of small mussels during the Middle Period. This suggests evidence for overpredation on mussels around the time that people adopt more fish and sea mammal in their diet. As these alternatives become more prominent, mussel size recovers, although never to the average size reflected early in time. Average size at CA-SRI-135 varies from 43.1 to 46.6 mm.

This study is directly applicable to understanding settlement distributions associated with initial movement into a new geographic region. The IFD has been shown to be effective at understanding settlement distribution and expansion over time (Codding and Jones, 2013; Yaworsky and Codding, 2018; Harvey, 2019; Weitzel and Codding, 2022; Jazwa and Collins-Elliott, 2021; Jazwa and Jazwa, 2022; Miller and Carmody, 2022). However, Jazwa et al. (2016b) highlighted limitations to the IFD for low population density. In the example given here, we argue that there is no benefit to small groups of people targeting the best habitats with the greatest potential for resource return. They can still individually act in a way that maximizes their return within their habitat (i.e., by targeting California mussels), but will have a comparable return at all habitats that meet a minimum threshold rank. In doing so, they might choose to distribute themselves on the landscape in such a way as to prioritize other traits typically excluded in ranking habitat suitability per the IFD. On SRI, prevailing winds and storms generally come from the northwest, and the central ridgeline of the relatively rugged island blocks the south side from them. Even during calm weather periods in the summer, the authors have numerous anecdotal observations of days in which the north coast is overcast and cold, but the south side is warm and sunny. Such conditions are less important during periods of resource stress when it is necessary to target habitats with the most abundant resources, but all else being equal, such factors may be considered. Considerations like viewshed (e.g., Lake et al., 1998; Jones, 2006) may also be relevant.

As important, a common assumption of the IFD is one of complete knowledge of the region under consideration, including sufficient information to accurately appraise and compare relative suitabilities for all habitats (Winterhalder et al., 2010: 471; Assumption f). This may not necessarily be the case, especially in a large geographic area, and a settling population may be willing to settle down in one of the first habitats that they encounter that is good enough to provide the subsistence resources they need in quantities that are more than they can consume, even if other, nearby habitats allow for greater resource maximization (Haas and Kuhn, 2019). In such cases, early settlement decisions may have been conditioned by factors such as proximity to previously settled regions. On the NCI, this is less likely to be relevant as the islands are relatively small and it is likely that settlers would be able to quickly assess all potential habitats. However, when considering the initial peopling of the Americas south of the Cordilleran Ice Sheet, for example, people would have first encountered habitats along their route rather than more distant interior habitats that may have had higher resource return. This requires a relaxation of the assumption of complete knowledge in cases in which such information is not available.

While we have chosen to model human settlement patterns for low population densities as a variant of the IFD with modified predictions, the behavior we describe here is consistent with satisficing rather than maximizing models (Bordley and LiCalzi, 2000; Schmidtz, 2004). Satisficing economic models are target- as opposed to maximization-oriented and have been shown to mediate decision making problems relating to risk, uncertainty, and incomplete knowledge (Halstead and O‘Shea, 1989; Low, 1990; Goland, 1991; Winterhalder and Leslie, 2002; Morgan, 2009), as may have been the case for early settlers on the NCI. The use of a formal satisficing model at low population densities that gives way to a standard IFD model once a population threshold has been met could provide insight that our approach might not.

Finally, our study is limited in that much of the land that would have been on the coast of Santarosae during the Paleocoastal period has since been submerged by marine transgression. While an increasing number of Paleocoastal sites have been found across the NCI (Erlandson et al., 2011, 2015, 2016, 2019, 2020; Rick et al., 2013; Jazwa and Rosencrance, 2019; Braje et al., 2020), most are inland. While several coastal sites have been found, CA-SRI-347 at the mouth of Trancion Canyon provides the best evidence for permanent settlement. We cannot say for certain that the patterns we observed in this study are characteristic of the full phase of early occupation, but rather the end of the early, low-density phase of human presence on the NCI. Regardless, we can say that population density was low enough at the very beginning of the Early Period that settlement should be focused on the highest-ranked habitats under the standard formulation of the IFD, which it clearly was not. Our study consequently provides insight into early settlement decisions at low population density when there is little or no resource stress, regardless of whether or not it represents the precise settlement distribution of the very earliest settlers of the NCI.