Considering that the instrumental climate record covers a period of about a century, it becomes necessary to use paleoclimatic records/models to explore the stability of the climatic variability and investigate the robustness of the teleconnections of the past. It is important to identify if the observed patterns in the current period persist over the last millennium when the changes in the orbital induced climate variations are negligible. In several studies on climatic causality crosscorrelation functions are used and the analysis is based on the relationship between atmospheric structures in pairs, a procedure that has several limitations in the elucidation of the network of possible connections. To mitigate these barriers, this work uses Partial Directed Coherence (PDC) and kernel nonlinear Partial Directed Coherence (knPDC) to allow the inference of the linear or nonlinear couplings between the climatological patterns, respectively. Connections between the two groups of climatic indicators in the last millennium were observed from 850 to 1850. The first group comprises the El Nino-Southern Oscillation (ENSO), Atlantic Multidecadal Oscillation (AMO) and Atlantic Interhemispheric SST Gradient (GTA) and the second, Antarctic Oscillation (AAO), El Nino-Southern Oscillation (ENSO), Pacific-South American (PSA1, EOF2) and QuasiBiennial Oscillation (QBO). The climate indices were computed from a weighted average set of climate model simulations of PMIP3, which represent simulations oriented to the past climate in the climate projection models of CMIP5. For the first group, no significant results were observed on the low-frequency band, observing only linear relationships between the Pacific and Atlantic Oceans. For the second group, the causal analysis point to linear relationships between ENSO↔AAO, and nonlinear between ENSO↔PSA in the low and high band and QBO↔AAO, QBO↔ENSO and QBO↔PSA in the low-frequency band. In summary, the results indicate a higher nonlinear connection between low-frequency phenomena.