This paper examines the causal relationships between investor attention and market return， using the residual bootstrap Granger non-causality test and a fixed-size rolling-window estimation approach. After comparing multiple indicators， we choose Hexun Attention Index as proxy indicator for investor attention. In order to retrospect the historical data， we develop a VBA program for web data extracting， and the final data sample covers the period from March 10，2009 to October 31, 2013. The full-sample bootstrap-LR test suggests the existence of a unidirectional causality running from market return to investor attention. Whereas， the parameters stability tests unveil that both the short-run and the long-run relationships between investor attention and market return estimated by full-sample data are unstable over the sample period， which in turn suggests the results of full-sample causality tests are unreliable． Hence， it motives the application of bootstrap rolling-window estimation to revisit the dynamic causal relationship between the two variables. The results indicate that investor attention has a significant effect on market return among some sub-samples， but over time such causality connection becomes weak. While market return has a great impact on investor attention in the rolling-window model， which is similar to the full-sample tests. Especially during 2009 and 2010， a vast majority of the sub-sample test results show that the market return is Granger cause of investor attention. Meanwhile， we average the estimated coefficient of VAR model of all sub-samplesto measure the mutual impact degree between investor attention and market return. The results confirm that the effect on market return by investor attention is negative， namely， high investor attention will lead to low market return. And market return has a significant positive effect on investor attention. At last， we adopt Hurst index to analyze the possible reasons behind the change of dynamical relationships between investor attention and market return. We find that it is the improvement of market efficiency that cripples the negative influence of investor attention on market return. It can be seen that predicting market trends and getting excess returns via investor attention index will gradually be impossible in the future. What’s more， with the volatility of the A-share market falling， the positive effect of market return on investor attention is also fading.  The creative points of this paper lie in: （1） using statistical data scraped from large financial website as an indicator of investor attention， which can minimize the measure bias; （2） the application of rolling-window estimation approach allows us to study the dynamic relationships between investor attention and market return. On the other hand， without establishing a dynamic window model based on listed companies’ data and shorter sample span limit the depth of this study．