The Wintu and Their Neighbors

Christopher Chase-Dunn and Kelly M. Mann

Appendix 2:

An analysis of projectile point data from the Sacramento River Canyon

This is an analysis of propoint data from Basgall and Hildebrandt (1989) to study obsidian source differences by location in the Sacramento River Canyon.

Tuscan obsidian is a particular type of volcanic glass that comes from a delimited area near Backbone Ridge (Seaman's Gulch) not far from the heartland of Wintu territory. Ethnographers have long asserted that the boundary between Wintu-speakers and the Okwanuchu-Shastans in the Sacramento River Canyon was just north of Pollard Flat where North Salt Creek joins the Sacramento. If the boundary between the Wintu and the Okwanuchu was there, and if the Wintu had been pushing north up the Canyon into Hokan-speaking territory:

• we should find a difference in the ratio of Tuscan to Medicine Lake obsidian as we proceed north, with Tuscan becoming rarer as we go north, and perhaps with a big difference at the boundary. (Unfortunately all the sites studied by Basgall and Hildbrandt are below North Salt Creek, though Pollard Flat is very close.)
• If Wintu incursion toward the north was proceeding over time we should also find more Tuscan obsidian later than earlier.

The above follows because the Tuscan sources (near Backbone Ridge) are well within Wintu territory while access to the Medicine Lake highlands (Northeast of Mount Shasta) requires that the Wintu cross through Shastan lands or trade for obsidian with their enemies.

Bill Hildebrandt provided four D-Base files containing the propoint data listed in Basgall and Hildebrandt (1989: Vol 2 Appendix J) -- one file from each site they studied. We converted these into a single SPSS file.

Variables that we added:

SITE = SHA-1175, SHA-1176 (both at Vollmers), SHA-476 (at La Moine), SHA-1169 at Pollard Flat.

LOCATION = Vollmers, La Moine, Pollard Flat

DISTANCE (from Seaman's Gulch in kilometers, the way the crow flies)

Vollmers= 38 km.; LaMoine= 42 km.; Pollard Flat= 45 km.

Then we merged the four files. The total number of projectile points is 1321.

The number of points for each LOCATION: Vollmers = 318 points; La Moine = 599 points; Pollard Flat = 404 points

Variables we recoded:

CHEM= chemical sourcing (x-ray florescence).

0= not sourced; 1= Grass Hopper Flat/Lost Iron Well; 2= Tuscan; 3 = Borax Lake; 4 = Glass Mountain; 5 = Buck Mountain; 6= Railroad Grade; 7 = Unknown; 8 = East Medicine Lake; 9 = Harris Flat.

We computed CHEM1 lumping GHF/LIW, East Med Lk, Railroad Grade, and Glass Mountain together into "Medicine Lake Highlands".

Medicine Lake Highlands N = 620
Tuscan N = 81
Borax Lake N = 1
Buck Mountain N = 5
Harris Flat N = 1
Unknown N = 1
Not Sourced N = 612

HYD -8.8= not cut for hydration, declare as missing value; Why so many 0.0 readings? What do these mean?

HYD1 = HYD categorized into 4 approximately equally-sized groups of points.

HYD3 = Medicine Lake hydration converted into Tuscan scale using formula on p. 198 in Basgall and Hildebrandt, Vol 1.
Tuscan microns = 0.670 GF/LIW/RS microns - 0.296.

HYD4 is HYD3 divided into categories :

1 = .0 to .7 = present to 565 BP N = 70
2 = .701 to 1.4 = 566 BP to 1280 BP N = 74
3 = 1.41 to 2.4 = 1281 to 2710 BP N = 182
4 = 2.41 to 3.4 = 2711 to 4515 BP N = 333
5 = 3.41 to highest= 4516 to very old N = 262

400 points were not cut for hydration.


SERIES (called "Class" in B and H, Appendix J)

12 point types. This is a string variable.
SPSS uses only the first 8 letters, thus collapsing some of the categories, e.g. Gunther A, B and C become Gunther.

We call our variable (based on SERIES) PTYPE. There are 231 Gunther points, 7 Desert Side-Notched, 411 Clikapudi, 88 McKee, 24 Pollard, 270 Squaw Creek, and 50 Widestems. Gunther points are sometimes considered a time-marker because these small points were used on arrows, and appear with the coming of the bow.

location x chem1
location x chem1 x ptype
location x chem1 by hyd4
chem1 x hyd4
chem1 x hyd4 x location
ptype x hyd4

Results

Location by Chem1 using only Tuscan and Medicine Lake points. n = 694


% tuscan by location


Location	% tuscan	N
Vollmers	12.1%	206
La Moine	7.2	319
Pollard	19.3	169


Overall Chi Squared Sig. 0003.; Diff between Vollmers and La Moine Sig. .08; Diff between Vollmers and Pollard Sig. = .07. Diff between La Moine and Pollard Sig. = .0001.



Controlling for Ptype = Gunther points only


Location	% Tuscan	N
Vollmers	24.4	41
LaMoine	7.1	56
Pollard	18.5	27


Overall Sig. .06; Diff between Vollmers and Pollard Not Sig. (.80). Diff between La Moine and Pollard Not Sig (.23).
(but Gunther points are not good time markers for the most recent period in this context -- see below)

Using HYD1, which is HYD uncorrected for source, divided into approximately equal size categories.

Hyd1= 1 (most recent) 0.0 to 2.7 microns, N = 225


Location	% Tuscan	N
Vollmers	25.5	51
LaMoine	20.7	82
Pollard	35.7	28

Not. Sig.



Hyd1 = 2 2.8 to 4.1 microns, N = 221


Location	%Tuscan	N
Vollmers	15.4	52
LaMoine	6.4	78
Pollard	43.3	30

Sig = .0000

Using HYD5, which is based on HYD uncorrected for source, but divided into categories which contain the same number of cases as HYD4 (in order to compare better with HYD4 results below).

HYD5 = 1 (most recent) 0.0 to 1.2 microns, N = 45


Location	% Tuscan	N
Vollmers	36.4	11
LaMoine	13	23
Pollard	45.5	11

Overall Sig. level = .0955


HYD5 = 2 , 1.3 to 2.2 microns, N = 62


Location	%Tuscan	N
Vollmers	26.1	23
LaMoine	33.3	33
Pollard	50	6


Overall Sig. level .53 (not sig.)

Using HYD4 (corrected for source diffs in hydration rate and categorized (see above)

Hyd4 = 1 (n= 52)


Location	%Tuscan	N
Vollmers	15.4	13
LaMoine	3.6	28
Pollard	44.5	11


Overall Sig. .0049; Diff between Vollmers and Pollard Not Sig. (.24). Diff between LaMoine and Pollard Sig. .0056.

Hyd 4 = 2 (n=65)


Location	%Tuscan	N
Vollmers	16.0	25
LaMoine	6.9	29
Pollard	11.1	9


Overall not sig.(.57) Diff between Vollmers and Pollard, Not sig.
Diff between LaMoine and Pollard Not Sig.

Lets cut time a little differently, looking at the most recent cases separately:
Only looking at cases which have score of .0 on Hyd3 (n= 23)
%Tuscan N
Vollmers 40 5
La Moine 12.5 8
Pollard 50 10


Location	%Tuscan	N
Vollmers	40.0	5
LaMoine	12.5	8
Pollard	50.0	10


cases which have gt .0 but lt 1.4 on Hyd3 (n= 91)


Location	%Tuscan	N
Vollmers	9.4	32
LaMoine	4.1	49
Pollard	10.0	10



Lets look at change over time.

HYD4 by Chem1 (all sites)


Hyd4	%Tuscan	N
0-565 BP	15.4	52
566-1280 BP	11.1	63
1281-2710 BP	11.6	164
2711-4515 BP	9.2	292
4516-old	15.0	80



Does this pattern vary by location? Yes.

Vollmers


Hyd4	%Tuscan	N
0-565 BP	15.4	13
566-1280 BP	16.0	25
1281-2710 BP	9.3	54
2711-4515 BP	8.3	12
4516-old	15.0	80


LaMoine


Hyd4	%Tuscan	N
0-565 BP	3.6	28
566-1280 BP	6.9	29
1281-2710 BP	12.6	87
2711-4515 BP	4.8	147
4516-old	8.0	25


Pollard


Hyd4	%Tuscan	N
0-565 BP	45.5	11
566-1280 BP	11.1	9
1281-2710 BP	13.0	23
2711-4515 BP	16.9	59
4516-old	20.9	43


What does this mean? There is almost always more Tuscan obsidian at Pollard . This is rather wierd because Pollard is the most distant location from the source of the Tuscan obsidian. But the big difference in the most recent time period is based on a small sample of 11 points and so that could be an aberration.

When we use HYD5, the uncorrected measure of hydration, the results are more encouraging for my hypothesis about increasing Tuscan use over time, but this is an artifact of the slower rate at which Tuscan obsidian hydrates. This demonstrates the importance of correcting for source-specific hydration rates.

HYD5 by Chem1



Hyd5	%Tuscan	N
0-565 BP	27	45
566-1280 BP	32	62
1281-2710 BP	21	129
2711-4515 BP	4	230
4516-old	2	185



Lets try looking at change over time by using point types as time markers. The tables below show that Gunther points are not good markers for the most recent time periods, but we can try using point types as time markers anyway. I will order the point types in the way indicated by the table below that uses HYD4 (corrected hydration) : Gunther, Clik Cor, McKee, Pollard, SC Contr and Widestem.



Point Type	%Tuscan	N
Gunther	15.3	124
Clik Cor	10.6	132
McKee	0	53
Pollard	6.3	16
SC Contr	16.1	149
Widestem	20	30

---




Conclusions:

Certainly the predictions made above were not borne out. If anything the spatial relationship is the reverse of what was predicted. Pollard Flat, the northernmost location and the one most distant from the Tuscan source seems to always have a greater proportion of Tuscan propoints. La Moine, the intermediate location, almost always has the fewest Tuscan, and Vollmers, the nearest to the Tuscan source and the closest to the centers of Wintu territory, usually has a proportion of Tuscan points which are intermediate between the other two. These differences are based on the use of hydration measurements corrected by source. Use of uncorrected hydration measurements (HYD1) produces the same pattern for the most recent time period.


A slightly different result is obtained using Gunther points as time markers for the most recent period, with Vollmers having the highest % Tuscan, Pollard 2nd and La Moine least, but this is suspect because Gunther points are not a clear time marker for the most recent period in the Sacramento River Canyon. Also the differences between the locations were not statistically significant.

What about changes in Tuscan usage over time? There do not appear to be significant differences across time periods when we use corrected hydration measurements. The increase in % Tuscan from 11.1% to 15.4% in the transition from HYD4-2 to HYD4-1 (all locations) is not replicated when we look at individual locations. Pollard Flat accounts for most of the increase, and this result is based on a Pollard Flat N for HYD4-1 of 11 points.

Using point types as time markers to examine change in Tuscan use over time shows a curious pattern of early high Tuscan use, little Tuscan in the middle, and increasing Tuscan use in late point types. I tend to trust the hydration measure of time more than the point types because of the results shown below. Most point types are spread widely across time and are poor time markers.

It may be that there are no true significant differences across these locations or across time. The differences between locations are sometimes statistically significant, especially between La Moine and Pollard Flat. Though this difference is significant at the .0056 level for HYD4-1, this is based on only eleven points from Pollard Flat. Part of the problem may be due to the rather small sample of recent points from Pollard. The difference between LaMoine and Pollard for HYD4-2 is not statistically significant.

It would be helpful to have a larger sample of recent points from Pollard Flat. Is it possible to do hydration and sourcing on those points which were not studied from Pollard in order to increase the size of the sample? There were 349 points which were neither cut for hydration nor sourced. Seventy-five of these were from Pollard Flat. How much would this cost?

If such a study were to confirm the a big difference of the kind indicated between La Moine and Pollard Flat, with Pollard having more Tuscan than LaMoine, how could we account for this? Obviously propinquity and transport costs would work against this outcome. What if there were an ethnic/linguistic boundary between La Moine and Pollard with lots of trade across the boundary. It is possible that groups on either side of the boundary would prefer the kind of obsidian that they had less access to, and that this would result in the differences? But if this were true, why would the use of Tuscan be greater at Vollmers than at La Moine, and why should this apparent pattern persist over 5000 years when ethnic boundaries probably changed greatly over that period?


Elaine Sundahl (1982:95,166) has found only one projectile point made from Medicine Lake Highlands obsidian out of sample of thirty points found at two sites in the Redding area. Twenty-eight others were Tuscan. This implies that Wintu in the Redding area were using very little non-Tuscan obsidian. It would be helpful to have data from studies intermediate between the Sacramento River Canyon and Redding to help map out the sourcing patterns. Thus the broader spatial comparison between Redding and the Sacramento River Canyon does seem correspond with our original arguement about increased Tuscan usage in the South.


Sundahl contends that the Wintu did not use much non-Tuscan obsidian. She thinks that the Sacramento River Canyon was not Wintu territory until very recently. Whether or not that is true, the sites studied by Basgall and Hildebrandt were determined largely by the rerouting of I-5, and probably do not correspond with the locations of major riverine villages, except perhaps for the La Moine site. The freeway route is quite high above the river, while most of the village sites were on flats rather close to the river. Since the Wintu were riverine people the site selection may better represent older, pre-Wintu inhabitants of the Canyon.


---


Checking our measures of time -- hydration and point type.

PTYPE by HYD5 (hydration uncorrected for source diffs)



Hyd5	Clik Cor	Gunther	McKee	Pollard	ScContr	Widestem
1(recent)	9 20%	16 36%	1 2%		8 18%	1 2%
2	14 23%	28 45%	2 3%		2 3%	2 3%
3	25 19%	58 45%	5 4%	1 1%	12 9%	6 5%
4	56 24%	13 6%	17 7%	6 3%	48 21%	11 5%
5(oldest)	21 11%	2 1%	25 14%	9 5%	67 36%	6 3%


PTYPE by HYD4 (corrected for source diffs in hydration).



Hyd5	Clik Cor	Gunther	McKee	Pollard	ScContr	Widestem
1(recent)	15 21.4%	22 31.4%	3 4.3%	1 1.4%	9 12.9%	1 1.4%
2	12 16.2%	47 63.5%	3 4.1%	0	0	1 1.4%
3	50 27.5%	55 30.2%	4 2.2%	1 0.5%	13 7.1%	4 2.2%
4	61 18.3%	31 9.3%	33 9.9%	13 3.9%	86 25.8%	16 4.8%
5(oldest)	36 13.7%	6 2.3%	18 6.9%	3 1.1%	94 35.9%	13 5.0%


Here is what this looks like when uncorrected Hydration measurements are divided into four approximately equally sized categories:


Hyd5	Clik Cor	Gunther	McKee	Pollard	ScContr	Widestem
1(recent)	42	109 48.4%	6	2	18	5
2	65	43 19.5%	6	1	26	7
3	51	7 2.6%	27	10	66	13
4(oldest)	16	2 1.0%	22	5	92	10


Gunther points are not a good time marker for the most recent period. When HYD1 is used (uncorrect hydration that has been divided into four categories of about the same size) Gunther points look like a strong measure. But this result is confounded when we look at results using HYD4 or HYD5. Both of these make finer distinctions within the most recent time period, and they show that Gunther points decline as a percentage of all points in the most recent period. This result is not dependent on the correction we made for hydration source, but is rather due to the way in which hydration measurements are grouped.
There are differences in the HYD4 and HYD5 tables which are important in their implications for the use of point types as time markers.We have used the corrected hydration table (HYD4) to order point types for our study of change in Tuscan use over time (see above).