The Proceedings of the Eighth International Conference on Creationism (2018)

described by Baillie (2015, p. 95), “If two ring patterns A and B definitely matched (without doubt), then any third pattern C which matchedAmust also match B at a unique position . If it did not, then the implication had to be that C was incorrectly matched against A. As more trees were matched, each new example could be checked against A, B, C, etc.” (Emphasis is in original). However, such reciprocity in crossmatching is not the sole property of correct crossmatches, and, moreover, can occur even among series that flout virtually all other dendrochronological criteria. [This is deliberately shown in Figure 5. The initial overlap (40 years) is much too short to start a chronology. The t-values are closer to the “floor” of 3.5 (one below it) than to the “ceiling” of 7.0, and, moreover, based as they are on Cofecha, some of the higher values may be inflated. The seven reciprocally-crossmatching series generally fail all the gateway statistics. Yet, despite all these fatal liabilities, all seven series, each assigned to very different ages in TRN, all reciprocally crossmatch at the same point.] The reciprocity of pair-crossmatches (e. g, Figure 5), especially large numbers of them, is best shown in a matrix. This is what is done by CDendro. 3. The Internals of Valid Crossmatches: The Block Test Thus far, attention has been devoted exclusively to the externals of the prospective crossmatch (the overall t-value for the overlapping sequences). To be accepted as a valid, the internals of the prospective crossmatch must also be satisfactory: All components of the overlapping sequences should match acceptably. That is, individual segments (blocks), of one series against the other, or of one series against the master chronology, should not fall below a specified r-value. This is called the block test(*), which in practice is the wrstblk (worst block) test. In this paper, 50-year segments (blocks), successively lagged 10 years, should not fall below r=0.3. (However, some authors use less stringent criteria, such as 50 years lagged 25, while others use stricter ones, e. g, 30 years lagged 10.) It is reasoned that a false prospective crossmatch may fortuitously have an acceptably-high overall t-value, but it is less likely that its variance will also fortuitously be distributed sub-evenly within. I have examined 44 high (t≥6.0; OVL≥80 years) FIN false pair- crossmatches in CDendro, and have found that about two-thirds of them fail the 50 years lag 10 wrstblk of r≥0.3. An example of one that does is shown in Figure 3. So do nearly all the series in Table 2 (failures being slight ones). A potentially-false crossmatch is not the only reason for preferring the exclusion of a series that fails the block test. An otherwise- qualifying series that has a poorly-matching internal segment is not a good choice for enhancing the common variance that is the main reason for building a master chronology in the first place. However, the rejection of prospective crossmatches that contain low-matching segments is not hard-and-fast, as it is not unusual for a tree to go through decades of suboptimal response to the annual climatic signal. In fact, depending upon the judgment of the individual dendrochronologist, prospective crossmatches with even multiple low-matching segments can still be accepted as correct (e, g., Matheus 2017). The foregoing extends to low-matching- segment-containing series against the master chronology, which is further aggravated whenever the sample depth* is relatively low, and there is insufficient common variance to “iron out” some of the “misbehaving” segments. But how often is the block test allowed to “slide” in the long chronologies? I used Cofecha quality control to identify series that contain flagged segments (defined as 50-years-lag-25-years Woodmorappe ◀ Tree-ring chronology shortening via disturbances ▶ 2018 ICC 657 Figure 3. Screenshot of the CDendro Screen Displaying a Convincing But False Crossmatch of Two Series, From the Finnish Long Chronology (FIN), That, According to Their Placement in the Master Chronology, Had “Lived 3,997 Years Apart”. The visual fit of the curves, especially at the all-important narrow rings, is good, and the internal consistency of the crossmatch is verified by the block test which shows consistently-acceptable r-values for each 50-year segment lagged 10 years: 0.50, 0.40, 0.50, 0.46, etc. The OVL is 149 years; the correlation of the two series according to P2YrsL is r=0.48 and t=6.6; the offset is 35 years.