Introduction to Archaeology: Class 5
Absolute dating: Tree rings and radiocarbon
ã
Copyright Bruce Owen 2002
- Two kinds of dating: relative and absolute
- Relative dating
puts things in order, older to younger, without specifying dates in years; we'll look at these later
- Absolute dating
gives ages in years.
- In theory, this is better than relative dating, because we know both how old things are, and can put them in order
- Unfortunately, most "absolute" dating methods give slightly fuzzy dates (radiocarbon dates are usually plus or minus 40 years or more), so sometimes we can get the order more precisely by lower-tech, relative methods.
- also, most absolute dating methods are expensive and involve sending samples to a lab and waiting for the results
- Some kinds of absolute dates:
- Historical dates
(coins, dated inscriptions, etc.)
- not as simple as you might think
- for reasons that apply to all kinds of absolute dates, not just historical ones
- the key is to have a clear understanding of archaeological associations of artifacts, features, activities, and dates
- Thomas throws in issues about associations while covering other things, but I think it is important to deal with them head on, right away
- and historical dates give us a clear, simple case in which to do it.
- say we excavate a historical burial and find a coin in it dated 1827.
- was the person buried in 1827?
- First, we have to feel sure that the coin (or any absolute date) is truly associated with the rest of the burial activity that we want to date
- that is, that the coin got into the burial as part of the burial event, not some other time
- we should try to rule out the possibility that the association is false due to disturbance that introduced artifacts after the event of interest
- maybe the person was buried back in 1500
- and in 1830, someone looking for treasure dug into the burial and accidentally dropped the 1827 coin into it.
- or in 1830 a gopher tunneled through the burial, and someone accidentally dropped the 1827 coin down the hole
- we can try to rule out disturbance by carefully observing the soil and the positions of the artifacts while excavating
- that is, we can say "we looked carefully for evidence of rodent burrows but did not see any"
- disturbance can introduce artifacts of any age
- the example above has an artifact that is much younger than the burial falling down the gopher hole into to
- but it could just as well be a paleolithic spearpoint that falls into the gopher hole
- that is, something much older than the burial
- the point is that an unrelated artifact gets added to the context after the fact
- such artifacts are not really associated with the event of interest; they have nothing to do with it at all, and tell us nothing about when the event happened
- we should try to rule out the possibility that the association is false due to redeposition of earlier material
- maybe the person was buried in 1900
- but when they dug the grave, they dug through a layer of garbage from the 1830s that included the coin
- and when they filled the grave, they redeposited the old coin right next to the casket
- again, we can try to rule out redeposition by observing carefully during excavation, and by lab analysis
- we can say "the surrounding soil was free of artifacts, charcoal, etc., so contamination of the burial by redeposited artifacts is unlikely."
- that is, we should not just assume that things found close to each other necessarily result from the same event
- we have to look carefully to rule out the possibility of disturbance or redeposition
- we usually cannot be absolutely positive
- some kinds of closed contexts (also called sealed contexts) give us confidence that the objects really got in there at the same time
- because there is no way for anything else to have gotten in
- for example, if we found the tomb or casket intact, we can be pretty sure that everything inside it really is contemporary
- Assuming that the coin was deposited at the time of the burial, it still does not tell us the date of the burial
- because the event that it dates is not the event we are interested in
- the date on the coin tells us when the coin was made
- not when it got into the burial
- so even though the coin is well associated with the burial, the date is not
- this is exactly the same problem that Thomas deals with extensively when discussing the dating of the Oseberg ship
- they couldn't just date the wood in the ship, because that date is not closely associated with the event of the burial
- they chose to date timbers that were apparently cut specifically for the burial chamber
- these dates are well associated with the burial event itself
- if the association is secure, what the coin does tell us is that the person was buried some unknown time after the coin was minted in 1827
- 1827 is the date after which the burial must have taken place
- this kind of "date after which" is called a terminus post quem
- these are extremely useful, but be careful: we don't know how long after...
- dendrochronology
(tree ring dating)
- Low tech, but the most precise method there is
- Most trees grow by adding one layer or "ring" of wood per year: a low-density, light-colored part in the rainy season, and a high-density, dark-colored part in the dry season
- The thickness of the rings varies depending on the climate each year
- a complex combination of rainfall, temperature, sunlight
- If you count inwards from the bark of a recently felled tree, the widths of the rings are a record of the climate of each year back to when the tree sprouted
- Actually, they usually use narrow cores drilled out of the tree, so they don't have to cut old trees down for this
- Any given period of years has a unique pattern of ring widths
- If you have a piece of wood with numerous rings, you can match its ring width pattern to the old tree and tell exactly which years those rings grew in
- The pattern can be extended back further into time by finding older logs that have ring width patterns that overlap
- this is extremely accurate - to the exact year
- someone must create a separate master sequence for each region, sometimes even for different types of trees
- this is a very time-consuming, expensive process
- and depends on both luck and persistence to find samples that overlap enough and leave no gaps
- range of dating:
- varies by region; up to 8000 years ago in a few places like northern Europe for Irish oaks and the US Southwest
- sometimes the best we can do is a "floating chronology"
- that is a tree-ring sequence that does not extend to a known date
- pieces of wood that grew during the period covered can be precisely dated relative to others, say "tree A was felled 32 years before tree B"
- but the starting date of the whole sequence is unknown
- the method cannot be used in all regions
- some areas have no suitable trees
- or the climate does not vary enough from year to year
- or the microclimates vary so much from place to place in the region that no single sequence would work
- or the trees are irrigated, so their ring widths reflect a combination of climate and irrigation activities
- this is the problem with most trees in coastal Peru
- including colonial olive trees that I collected wood from
- but also prehistoric trees that typically grew alongside canals
- or the work to establish a master sequence simply has not been done yet
- the method cannot be used at all sites
- it requires relatively large chunks of wood with quite a few rings
- if you don't have beams or posts at your site, you generally can't use this method
- the real trick, as with all methods, is ensuring a meaningful archaeological association of the date with an event that interests us
- we care about human behavior, not when a certain bit of wood grew
- dendrochronology is usually useful for logs that are preserved all the way out to the bark, or to the smooth surface just under it
- because it can then tell us when the tree died
- which is probably when someone cut it down
- which is probably near when it was used
- but:
- what if the tree died naturally, and was collected later for use?
- what if the tree was cut down and left for years to dry before it was used?
- what if a log in the building we want to date was not cut down for that building, but instead was salvaged from an old, abandoned structure?
- if squared beams or planks were cut from the log, or artifacts were made from small pieces of the log, an unknown number of outer rings have been removed
- so the growth of the wood in the object can be dated, but the tree might have lived many years after that before being chopped down
- even so, that gives us a date after which the artifact must have been made
- the object cannot be older than the tree rings present in it
- a terminus post quem
- this can help us bracket events in time, but we have to be careful not to confuse a terminus post quem with the actual age of the thing, since all we know is that it was made an unknown time after the terminus post quem.
- Radiocarbon
dating (carbon-14 dating, or 14C dating)
- Thomas's presentation of this is slightly confused; read it, but focus on the version posted on the class web site and discussed here
- Go through the online handout for an explanation of the method
- A minor wrinkle: two different estimates of the half-life of 14C have been used
- The original, early work used the "Libby" half-life of 5568 years.
- Later work produced a better estimate that is now universally used: 5730 years.
- So some early radiocarbon dates have to be recalculated to get a more accurate result.
- How is the measurement done?
- Two methods: conventional and AMS (accelerator mass spectrometer)
- Conventional
- pretreat the sample to remove contaminants
- burn the sample and collect the carbon dioxide gas (CO2) that is produced
- convert the CO2 through several steps to benzene (a carbon compound)
- put a measured amount of the benzene next to an instrument (basically a geiger counter) that responds every time it is hit by a beta particle (high-speed electron)
- every time a 14C nucleus decays, it emits a beta particle, and some of these hit the detector
- after letting your sample sit there for hours or days, the number of hits on the detector gives you a measure of how often 14C nuclei are decaying in the sample, which is proportional to how many are in there
- you do exactly the same thing, with the same setup, using a "modern" carbon sample
- (actually a standard with a known 14C content relative to the pre-bomb atmosphere)
- if the archaeological sample is emitting beta particles at, say, 1/4 the rate of the modern sample, it must contain 1/4 as much 14C
- from there, you can calculate how long it has been since the sample died
- in this case, two half-lives, or 11,460 years
- AMS
- Pretreat the sample to remove contaminants
- Burn the sample, collect the CO2, convert the carbon in it to a solid "graphite target"
- The target is mounted in a particle accelerator
- The accelerator makes carbon nuclei stream off the target in a narrow beam
- The beam contains a mixture of the three isotopes, reflecting the proportions of isotopes in the target
- The beam passes by some strong magnets, which bend the beam due to the electric charge of the nuclei
- the nuclei of the different carbon isotopes are each deflected to a different degree, due to their differing mass
- so the magnetic field splits the beam into three parts, each with nuclei of just one isotope
- a instrument is located in the path of each of the three beams that records the number of nuclei that strike it
- so the device literally counts how many nuclei of each isotope comes off the target during a given period of operation
- so it is easy to calculate the fraction that is 14C
- pros and cons
- AMS dating can use much smaller samples
- so it can be used when no large amounts of organic were found
- or on objects that you don't want to seriously damage by removing a big piece
- AMS dating is often more precise (smaller error estimate)
- AMS dates typically cost more, but they are getting more reasonable
- sample preparation
- both methods require that the sample be treated to eliminate possible contaminants: carbon that is older or younger than what we want to measure
- the methods vary depending on the material to be dated
- mechanical cleaning or sorting, often under a microscope
- often done by the archaeologist before submitting the sample to a lab
- chemical treatments that remove known kinds of contaminants that contain carbon, like humic acids from soils
- this requires physical chemists with a special background in radiocarbon issues
- usually done by experts at the lab
- Understanding the error term
- Both methods are based on a measurement of the amount of 14C present
- Like all measurements, these have some degree of uncertainty
- So radiocarbon dates come with an error term
- Like 500 BP ± 40
- The error term is the standard deviation (often called sigma, or σ) of the probability distribution (a "normal" or "bell" curve) of the estimated date
- The "500 BP" is the mean, or center, of that distribution
- The error term "± 40" is an indication of how wide the central portion of the probability distribution is
- The error term tells us that there is a 68.26% chance (not 67%, as Thomas says) that the true date falls in the indicated range (in this example, 460 to 540 BP)
- That still leaves almost a 1 in 3 chance that the date falls outside that range
- In order to be more certain, people sometimes double the error estimate (they give the "two sigma" error term)
- There is a 95% chance that the true date falls within this wider range
- That still leaves a 5% (1 in 20) chance that the true date is outside this range
- One way to reduce this uncertainty is run numerous dates
- There are statistical methods for combining multiple dates of the same event in order to narrow the range of uncertainty
- Comparing dates
- Because you don't know the actual date, but instead just a probability distribution of where it is most likely to fall, comparing dates can be complicated
- Say you date charcoal from a fire pit at 100 ± 40 AD, and a burned bone nearby at 150 ± 40 AD.
- is it possible that the bone was burned in the fire pit?
- or was the bone burned somewhere else at a later time?
- or could the bone even have been burned before the fire pit was used?
- You simply cannot know for sure
- But using statistics (or a computer program that does this), you can estimate how likely each of those scenarios is
- or how unlikely it is that they are not correct
- Combining dates
- Say you have 3 dates:
- 100 ± 40 AD
- 150 ± 40 AD
- 200 ± 40 AD
- They all come from the same excavated house.
- How long was the house occupied?
- From 100 to 200 AD?
- Not necessarily; there is a good chance that the earliest date is actually from before 100 AD, and/or that the latest date is actually from after 200 AD
- From 60 to 240 AD?
- Not necessarily; there is a good chance that the starting date is less than 40 away from the central estimate, or even on the plus side of it; same for the ending date
- There is also a smaller but real chance that one or both dates might be even further from the central estimate
- Worse yet, we might not have even sampled the whole period of occupation
- We didn't necessarily get samples from the very first and the very last moments of occupation
- Odds are that our samples fall somewhere within the period of interest, but not all the way at either end of it
- This is a complicated problem that many archaeologists are too intimidated to acknowledge
- a lot of otherwise competent archaeologists misuse radiocarbon dates because they don't understand (or want to deal with) the statistical problems involved
- nothing can be done about the problem of samples that don't cover the whole period of interest
- but if we have reason to think that they do cover the whole period, there are statistical ways to estimate the duration of the period and the uncertainty of the starting and ending dates
- the more dates, the less uncertainty in the estimates
- OxCal is the program I prefer for this kind of work; it is available free online from the Oxford Radiocarbon lab
- Calibrating
dates (dendrocalibration)
- Necessary because the amount of 14C in the atmosphere has not been quite constant, as noted in the online handout
- By dating specific tree rings of known ages, it has been determined what "radiocarbon age" corresponds to what "calendar age"
- This is presented as a "calibration curve", with radiocarbon age on one axis and calendar age on the other
- You essentially just read off the calendar age for whatever radiocarbon age you get from the lab
- But nature has been mean to us here
- The amount of 14C in the atmosphere has jiggled up and down on short-term basis
- So the calibration curve has not only a general trend, but also lots of local peaks, valleys, and flat spots
- That means that for some radiocarbon ages, there are several, even many, possible calendar dates that correspond to it
- That is, when the 14C content of the atmosphere was going down, organisms that died started off with less 14C than did others that died years before
- In severe cases, a recently dead organism could have less 14C (due to a drop in the atmospheric content) than a long-dead one that started with more and has not lost too much by decay yet
- That is, the recently-dead organism could give an older "date" than the long-dead one
- So some radiocarbon ages, regardless of how narrow the uncertainty is, can only indicate fairly wide spans of calendar ages
- There is no way around this problem, other than to use a different dating method
- So for certain periods of time, radiocarbon dating is not very useful for short-term chronology
- These "jiggles" in the calibration curve also complicate the uncertainty estimates
- Since they, too, have to be calibrated
- And this may shift them closer or farther from the central estimate
- Especially if they fall near a hump or flat spot on the curve
- So calibrated dates cannot be expressed as XXX ± YY years, because the error estimate on each side may be different
- Two solutions: just cite the range from the 1-sigma lower estimate to the 1-sigma upper estimate
- Or present a graphical representation of the probability distribution
- Calibration is generally done with a computer program
- Correcting for isotopic fractionation using the 13C correction
- Isotopic fractionation: the separation of nuclei according to their isotope (number of neutrons)
- So far, we have assumed that plants and animals incorporate 14C in exactly the same proportion as the atmosphere
- That is only approximately true
- In fact, there are very slight differences in the chemical properties of different isotopes
- Essentially, the heavier isotopes (with more nuclei) have more inertia, and actually move a little more slowly through chemical reactions
- So with every chemical step in a process like the formation of living tissues, the end product is slightly depleted in 14C compared to the chemicals available as inputs to the reaction
- a bit more of the 12C gets through the reaction, and a bit less of the 14C
- Over any series of chemical reactions, this increasingly depletes the amount of 14C in the end product, compared to the amount the reaction started with
- The amount of depletion depends on the particular series of chemical reactions
- So plants are slightly depleted in 14C relative to the atmosphere, and the animals that eat them and convert their carbon into other compounds are slightly more depleted in 14C than the plants
- It turns out that even different plants are depleted to slightly different degrees
- So if we measured the 14C in a living plant, it would be slightly less than the amount in the atmosphere
- it will seem that this living plant has already had some of its 14C decay away
- That is, we will get an erroneously old date on it
- and worse yet, the amount of the error is different for different materials, even different kinds of plants
- Example errors of different materials caused by isotopic fractionation:
- wood and charcoal require little or no correction (everything else is adjusted to correspond to average wood)
- Peat gives dates about 35 ±95 years too young
- Typical herbivore bone gives dates about 80 ±35 years too old
- some arid-adapted plants like papyrus give dates about 195 ±80years too old
- other arid-adapted grasses like maize (corn) give dates around 245 ±50 years too old
- animals who eat these plants (like maize) give dates around 375 ±35 years too old
- Fortunately, nature has given us a way to measure exactly how much depletion has occurred
- so we don't have to rely on the averages above, with their ranges of variation; we can find out exactly what the correction should be for any particular sample
- Because there are three isotopes of carbon, and two of them, 14C and 13C, are stable
- The difference in the rate of chemical reactions is directly proportional to the difference in the mass of the nucleus
- So, regardless of the reaction, the depletion of 14C (2 neutrons heavier than 12C) is exactly twice the depletion of 13C (1 neutron heavier than 12C)
- So all we have to do is measure the fraction of 13C in the sample and compare it to the fraction of 13C in some standard material
- The standard happens to be a kind of fossil which is plentiful (fossil belemnites from the Pee Dee formation)
- It does not matter what standard we use; it is just a reference to let us compare different materials
- If, say, the sample is 1% depleted in 13C relative to the standard, the sample must be 2% depleted in 14C for the same reason
- Knowing this, we can mathematically "add back" that amount of 14C
- that is, by measuring the amount of 13C in the sample, we can correct for the isotopic fractionation in any material
- We don't even have to know what the material is (what kind of plant, for example)
- The correction is automatically right, because whatever reaction rate differences affected the 13C must have affected the 14C exactly twice as much
- This correction is necessary, because:
- different materials have formed through different chemical pathways
- so they started off with different degrees of depletion of 14C when they died
- so they give dates that are incorrectly old to different degrees
- all of these can be fully corrected for if - and only if - the 13C content of the sample is measured and included in the age calculations
- otherwise, if you know the material, you can estimate a correction
- but there is variation within each material, so you have to increase the uncertainty of the adjusted date
- Marine carbon reservoir correction
- Organisms that live in the sea get their carbon from the CO2 dissolved in the seawater (the "marine carbon reservoir")
- This oceanic carbon has less 14C in it than the atmosphere does
- Because the oceans circulate slowly and are not perfectly mixed
- So water below the surface has not been in contact with the atmosphere for years, even centuries
- So some of the 14C dissolved in it has decayed away and has not been replaced from the atmosphere
- So marine organisms start off with less 14C than do terrestrial (land) organisms
- Thus they give radiocarbon dates that look older than they really are
- On a global average, the carbon in surface ocean water is about 400 years "old"
- So marine organisms look about 400 years older than they really are
- This can be corrected for if we know the depletion of the water they lived in
- Unfortunately, not all ocean water is the same
- Shallow, surface waters are only slightly depleted of 14C
- Deep waters are typically more depleted, because they have been out of contact with the atmosphere for longer
- In some places (like the coasts of California and Peru), currents cause deep waters to well up to the surface
- So fish and shellfish in those areas are even more depleted than they are in other areas
- So much so that they can appear to be 800 years old or more on the day they die
- In extreme cases, the error can be even worse
- So in order to date marine shell, fish bone, whale bone, etc., we have to first study the seawater the organism would have lived in
- And include the "apparent age" of the seawater in the calculations of the sample's age
- Actually, the "apparent age" is not widely used.
- Instead, a global average of 400 years is applied to all marine samples as a first approximation.
- When the age of the local seawater has been determined, you can get a better approximation of the date by including the local deviation from that average, which may be positive or negative.
- this local adjustment to the global average is called ΔR (delta-R, for "reservoir")
- What about animals or people who eat seafood?
- Since they get some of their carbon from these marine organisms that they eat, the animals or people are, in turn, depleted in 14C while they are still alive
- They will give incorrectly old radiocarbon dates
- The error will depend both on how "old" the local seawater is, and what fraction of their diet came from the sea
- This is adding a lot of estimated values to the date calculation, each with its own uncertainty
- It is usually better just to avoid all these problems by
- Not radiocarbon dating fish or shellfish
- Not radiocarbon dating samples of animals or people who might have eaten seafood
- Preferring to radiocarbon date terrestrial plant or animal material only
- Regional atmospheric carbon reservoir corrections
- Thomas simplifies when he says that the world's atmosphere is so well mixed (p. 73)
- in fact, it looks like there are minor but consistent regional variations in the 14C content of the northern vs. southern hemispheres, or even south Pacific islands like New Zealand vs. continental areas in the southern hemisphere
- these can be easily corrected for, once they are sufficiently well measured
- they only amount to differences of 50 radiocarbon years or less
- in many cases, people ignore these corrections
- but certain parts of the calibration curve can magnify the error to as much as 100 or more calendar years
- the local atmospheric reservoir affects all terrestrial materials equally, and apparently did not change much over time
- so it does not affect the order or spacing of dates in any one region
- it only affects how that sequence lines up with others on different continents
- this is not very important to most research questions
- The "old wood" problem
- Radiocarbon dating tells you when something died
- But wood may lay around for years, even centuries, before someone picks it up to use for fuel
- So a chronic problem with 14C dates is the "old wood" problem, in which wood charcoal from a fire pit is dated, but gives a date long before the true one, because the wood died long before the fire was burned
- With a single date, there is often no way to notice this
- With several dates from a single hearth, hopefully several different pieces of wood will be sampled. If there is wide variation in the dates, some or all of the wood is probably old
- With dates in a stratigraphic sequence, if the dates are in the same order as the stratigraphy, they are probably OK
- If not, some may be from "old wood"
- The best thing is to date things that are unlikely to have been hanging around for a long time
- I date a lot of prehistoric cloth, because it is made from wool that probably would not have been stored for long before the cloth was made, and the garments probably did not last more than a decade or two at the most in use
- Or date wood posts used in construction, which probably would not have been laying around naturally
Sample sizes
- depends on the method
- Conventional dating
- charcoal: typically 15-30 grams (a handful)
- wood: 25-100 grams (5 to 20 pencils)
- wool textiles: 50-100 grams (a large handful)
- bone: 300-500 grams (almost a pound!)
- AMS dating
- charcoal or wood: 5 mg (5/1000 of a gram!), (a few small pieces)
- wool textiles: 5 mg (postage-stamp size, or less)
- bone: 1 gram (one to several inches of a rib)
Age of the samples
- while some results have been gotten up to 70,000 BP or so, the Arizona NSF Radiocarbon Facility says they can measure dates only up to about 48,000 BP - and they are one of the leading labs
Collection and handling of samples
- Two main issues: association and contamination
- Make sure the archaeological association is good
- That is, that dating the sample will really date the event of interest
- May involve identifying the species of plant material, or careful study of the stratigraphy
- Avoid contamination with other archaeological material or modern carbon
- Don't drip your sweat on the sample!
- Don't drop cigarette ash or pencil lead (graphite) into the sample!
- But in fact, most sample pretreatment now strips that kind of stuff out
- It is best if there was no contamination to begin with, but it is rarely likely to be a real problem
- often said that samples should never be touched, and should be wrapped in aluminum foil (no carbon at all) inside a plastic bag
- again, this is not a bad idea, but is no longer seen as so important
- surfaces are removed in pretreatment
- polyethylene bags don't contaminate specimens
Cost per sample, with 13C measurement
- Conventional - Geochron lab
- standard: $300
- extended counting period (better precision or smaller samples): $400
- AMS at Arizona NSF Radiocarbon Facility:
- private museums, firms, individuals: $600
- research labs, universities: $400
- research done under an NSF grant: $200
Some take-home points:
- Radiocarbon dating works. No serious scientist doubts the method
- although it may fail in specific cases due to contamination of the samples, poor association, "old wood", or other problems
- a radiocarbon date tells you how long it has been since the tissue of a living thing died
- so only organic materials can be dated, that is, things that were once alive, like wood, bone, shell, leather, wool or cotton cloth
- unfortunately, organic materials are often not preserved
- but carbonized wood (charcoal) often survives even when other organic materials have decayed away
- so charcoal is the most commonly dated material
- of course, it is only found where there were fires
- inorganic things like pottery, buildings, stone tools, etc. cannot be radiocarbon dated themselves
- if you want to date an inorganic object, you have to date some organic material found together with it
- you have to destroy the material that you date, so you have to have enough organic material that you are willing to sacrifice a part of it
- it is not precise, that is, radiocarbon dates give a range, not a specific year
- like 550 AD ± 40 years
- the error term is one standard deviation (one sigma)
- which means that there is a 67% chance that the true date of the sample is between 510 AD and 590 AD
- Radiocarbon dates should always be corrected according to their measured 13C content
- Otherwise, different materials give differently biased dates, sometimes off by centuries
- Dates on marine material must be corrected for the marine carbon reservoir's difference from the atmosphere
- radiocarbon ages must be "calibrated" to make them correspond to calendar years
- most of the calculations are easily done using commonly available specialized computer programs
- dates are expensive, so we never have enough of them
- even with the best of dates, drawing the right conclusions depends on:
- making sure that the sample is truly associated with the event we want to date
- understanding the uncertainty and statistical issues involved in comparing and combining them correctly