This map shows speakers who have a monophthongal realization of /eh/ before /r/, as in Mary and Sarah. This map suggests that /ehr/-monophthongization is native to Eastern New England and the southern part of Western New England, which was settled from Plymouth and Rhode Island—perhaps as an original feature of all the initial southern New England settlements. This isogloss has just about the same gap in southwestern Massachusetts as TLN does, and just about the same unexpected wedge of New Hampshire and Maine. Since southeastern N.H. appears to have developed TLN independently from Western NE, for socially rather than phonologically motivated reasons, this suggests that /ehr/-monophthongization, which reduces the margin of security between merry and Mary, might be a necessary precondition in New England for TLN to spread throughout a region. There are isolated cases of TLN scattered across the map, but none of them outside the green isogloss of Map 3 managed to spread throughout a larger area.
This, then, allows us to explain the TLN gap in southwestern Mass. We can hypothesize that Western NE had had monophthongal /ehr/ as an original feature, but southwestern Mass. gained a diphthongal realization of /ehr/ through influence from neighboring portions of Eastern NE. (Since this is not a reversal of a merger, it is not unreasonable to suppose regional influence of this type, though we could not have argued that SW Mass. simply lost TLN through this type of neighbor influence.) Then, when TLN spread through Western NE, it did not take hold in the corner of southwestern Mass. in which Mary was diphthongal.