The city of Erie, Pennsylvania represents an anomalous case in the dialect geography of North America. According to all available historical records, it was linguistically aligned with the North in the early part of the 20 th century: the lexical data presented in Kurath (1949) and Carver (1987) locate Erie within most of the Northern isoglosses, and the phonological data presented in Kurath and McDavid (1961) show that Erie shared nearly all of its phonological features with the North and only a few with the Midland. However, recent research for the Atlas of North American English (Labov et al. 2006) shows that Erie is now a Midland city, and the two ANAE speakers from Erie show no traces of the Northern Cities Shift. Crucially, the two pivot points in the vowel system, as defined by Labov (1991), show clear Midland characteristics: short-a exhibits raising before all nasals, but not the general raising of the NCS, and both speakers have a complete merger of the vowels in cot and caught. Erie’s shift from being a Northern city to a Midland city is surprising given that the North/Midland boundary is the most clearly defined dialect boundary in North America today (Labov et al. 2006). Furthermore, it would not be predicted by dialect diffusion models that only take population and distance into account, such as Trudgill’s (1974) Gravity Model: Buffalo and Cleveland, the large Northern Cities along Lake Erie on either side of Erie are more populous and closer to Erie than Pittsburgh, the nearest large Midland city. The current study provides a more detailed characterization of Erie, and presents vowel measurements from seven Erieites, ranging in age from 25 to 60. I n general, the results confirm ANAE’s finding that Erie is aligned with the Midland. H owever, the vowels systems of the Erie speakers are different from the neighboring Midland speakers in two respects. First of all, /ow/ does not participate in the strong fronting that is characteristic of Pittsburgh/Western PA: only the youngest speaker (a 25-year-old female) shows an F2 value for /ow/ that is higher than would be expected for a Northern speaker. Furthermore, while all speakers clearly have the low-back merger, the phonetic realization of the resulting phoneme is unrounded and lower than the distinctly rounded and raised open-o of the Pittsburgh area. Thus, while Erie is clearly phonologically aligned with Pittsburgh, the two regions are not phonetically identical. This realignment with the Midland suggests that Pittsburgh has had a stronger influence on Erie since the middle of the 20 th century than either of the two large nearby Northern cities. Qualitative evidence from sociolinguistic interviews will be presented to confirm this and to show that Erieites have more contact with speakers from Pittsburgh than either Buffalo or Cleveland. Much of this contact stems from the popularity of Erie as a summer vacation destination for residents of Pittsburgh, evidenced by the fact that some Erieites refer to these summer vacationers from Pittsburgh as "mups" (from "come up"). It will be argued that this higher density of communication caused Erie to shift its phonological allegiance from the North to the Midland, and, consequently, that any model of dialect diffusion must take communication patterns into account in order to be fully explanatory.