Some links between linguistic meanings and markers change systematically and cyclically in what are known as "grammaticalization paths." Although the empirical observations that give rise to these characterizations are cross-linguistically robust, the representations and cognitive processes that support these shifts are still not well understood. This dissertation bridges that gap through the use of corpus studies and experimental methods, ultimately providing a characterization of the forces at play in semantic variation and change. The focus is on the Progressive-to-Imperfective shift. In this change, languages that have only one marker for "habitual" and "event-in-progress" readings recruit a new marker for the "event-in-progress" one. The two markers slowly undergo a categoricalization process, where each marker gets restricted to one reading, until the new marker generalizes to both readings and pushes the old marker out of the language. I look at this change in Spanish, since it currently presents two diachronically-related markers that convey these readings: the Simple Present and the Present Progressive. However, there is a many-to-many mapping between forms and meanings: both markers can express both readings, showing that a full categorical stage has not yet been achieved, though a generalization process is already underway. What are the contextual conditions that enable the use of each marker for each reading? Are these constraints related to the diachronic development of the markers? To address these questions, I propose that semantic variation and change are rooted in the structure of the meanings that participate in the shift and in the contextual conditions in which these markers are interpreted. On the basis of clear formal characterizations of the Progressive and the Imperfective, I argue that these meanings share a conceptual structure that allows for variation and change to occur. As for the contextual conditions, I propose that shared perceptual access between speaker and hearer is the relevant factor preventing full categoricalization. Shared perceptual access is a contextual means to attain perspective alignment, a key communicative goal grounded on two complementary cognitive capacities: Common Ground and Theory of Mind. While the former affords the speaker greater reliance on context, the latter forces her to be linguistically explicit. As for the generalization of the Present Progressive to "habitual" readings, I argue that it preferentially occurs in contexts that satisfy the presuppositional requirement of 'estar' (the auxiliary in the Present Progressive); that is, in contexts that entertain salient alternatives at which the proposition at issue does not hold. Given its semantic composition, the Present Progressive conveys the "habitual" reading and implicates a rejected alternative, maximizing context set restriction, increasing the informativity of the expression, and making it overtime preferable to the Simple Present, which conveys the "habitual" reading alone. These hypotheses are tested in three ways. A corpus study across three time periods confirms the role of shared perceptual access as the crucial contextual condition in the diachronic interplay of these markers. Acceptability-judgment tasks and self-paced reading studies also assess the role of these contextual constraints in three Spanish dialects (Central Peninsular, Mexican Altiplano, and Rioplatense). Results show that the Present Progressive is the preferred marker to express the "event-in-progress" reading (higher acceptability, shorter reading times), but the Simple Present can still convey it when speaker and hearer share perceptual access. This contextual boost disappears in Mexican Spanish, where participants reject the Simple Present regardless of context. For the "habitual" reading, the Simple Present is the preferred marker, but the Present Progressive can be used when the presuppositional content of 'estar' is satisfied by context. In Mexican Spanish, this contextual support is not needed. Thus, the data shows that Mexican Spanish is further along the grammaticalization path than its Rioplatense and Peninsular counterparts. Altogether the patterns observed across dialects are consistent with a model of semantic variation and change embedded in a communicative system, visible during real-time comprehension, and subject to identifiable contextual factors. On the one hand, linguistic markers compete to optimize Common Ground and Theory of Mind pressures, supporting each dialect's advancement in their own path of change. On the other hand, the "generalization" process is shown to be driven by the contrastive informativity strength of the combined lexico-semantic properties of the Present Progressive marker. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]