#valuecoding — Public Fediverse posts

I keep seeing studies (especially computational ones) citing the Ambrose et al. 2016 paper as proof that #HippocampalReplay is modulated by reward.

I don't think this result can be trusted...

1. The time spent at low-speed at reward sites wasn't controlled. Rats left the less rewarded sites much faster, and would probably move a lot more when less reward was present (not analyzed I think).
2. The amount of ripples or replay is analysed in count per second. I believe this means seconds of any kind of behaviour, not just rest or low speed. But #Ripples and the related #Replay only happens when the rat is quasi-immobile. At low speed you might have 1 replay per second, while when moving it will be 0/s.
3. A recent preprint (Mallory et al., 2024) suggests that after a few seconds at a reward location, reverse replay dominates. Ambrose et al. specifically show an increase of reverse replay for the x4 reward vs x1 reward. If rats indeed spent more time at the most rewarded location, they would be in reverse replay mode, fully explaining the results without the need to invoke any kind of value processing. Note: the reduced replay result was not specific to reverse replay for 1 vs 0, probably because the rats didn't stay immobile long enough to get some of those forward replays during the earlier reward period.

I might have missed something (if so let me know) but I am now completely unconvinced by the 2016 paper. What do you all think?

#Hippocampus #ValueCoding
Edit: added some more info after re-reading the 2016 paper

I keep seeing studies (especially computational ones) citing the Ambrose et al. 2016 paper as proof that #HippocampalReplay is modulated by reward.

I don't think this result can be trusted...

1. The time spent at low-speed at reward sites wasn't controlled. Rats left the less rewarded sites much faster, and would probably move a lot more when less reward was present (not analyzed I think).
2. The amount of ripples or replay is analysed in count per second. I believe this means seconds of any kind of behaviour, not just rest or low speed. But #Ripples and the related #Replay only happens when the rat is quasi-immobile. At low speed you might have 1 replay per second, while when moving it will be 0/s.
3. A recent preprint (Mallory et al., 2024) suggests that after a few seconds at a reward location, reverse replay dominates. Ambrose et al. specifically show an increase of reverse replay for the x4 reward vs x1 reward. If rats indeed spent more time at the most rewarded location, they would be in reverse replay mode, fully explaining the results without the need to invoke any kind of value processing. Note: the reduced replay result was not specific to reverse replay for 1 vs 0, probably because the rats didn't stay immobile long enough to get some of those forward replays during the earlier reward period.

I might have missed something (if so let me know) but I am now completely unconvinced by the 2016 paper. What do you all think?

#Hippocampus #ValueCoding
Edit: added some more info after re-reading the 2016 paper

I keep seeing studies (especially computational ones) citing the Ambrose et al. 2016 paper as proof that #HippocampalReplay is modulated by reward.

I don't think this result can be trusted...

1. The time spent at low-speed at reward sites wasn't controlled. Rats left the less rewarded sites much faster, and would probably move a lot more when less reward was present (not analyzed I think).
2. The amount of ripples or replay is analysed in count per second. I believe this means seconds of any kind of behaviour, not just rest or low speed. But #Ripples and the related #Replay only happens when the rat is quasi-immobile. At low speed you might have 1 replay per second, while when moving it will be 0/s.
3. A recent preprint (Mallory et al., 2024) suggests that after a few seconds at a reward location, reverse replay dominates. Ambrose et al. specifically show an increase of reverse replay for the x4 reward vs x1 reward. If rats indeed spent more time at the most rewarded location, they would be in reverse replay mode, fully explaining the results without the need to invoke any kind of value processing. Note: the reduced replay result was not specific to reverse replay for 1 vs 0, probably because the rats didn't stay immobile long enough to get some of those forward replays during the earlier reward period.

I might have missed something (if so let me know) but I am now completely unconvinced by the 2016 paper. What do you all think?

#Hippocampus #ValueCoding
Edit: added some more info after re-reading the 2016 paper

I keep seeing studies (especially computational ones) citing the Ambrose et al. 2016 paper as proof that #HippocampalReplay is modulated by reward.

I don't think this result can be trusted...

1. The time spent at low-speed at reward sites wasn't controlled. Rats left the less rewarded sites much faster, and would probably move a lot more when less reward was present (not analyzed I think).
2. The amount of ripples or replay is analysed in count per second. I believe this means seconds of any kind of behaviour, not just rest or low speed. But #Ripples and the related #Replay only happens when the rat is quasi-immobile. At low speed you might have 1 replay per second, while when moving it will be 0/s.
3. A recent preprint (Mallory et al., 2024) suggests that after a few seconds at a reward location, reverse replay dominates. Ambrose et al. specifically show an increase of reverse replay for the x4 reward vs x1 reward. If rats indeed spent more time at the most rewarded location, they would be in reverse replay mode, fully explaining the results without the need to invoke any kind of value processing. Note: the reduced replay result was not specific to reverse replay for 1 vs 0, probably because the rats didn't stay immobile long enough to get some of those forward replays during the earlier reward period.

I might have missed something (if so let me know) but I am now completely unconvinced by the 2016 paper. What do you all think?

#Hippocampus #ValueCoding
Edit: added some more info after re-reading the 2016 paper

I keep seeing studies (especially computational ones) citing the Ambrose et al. 2016 paper as proof that #HippocampalReplay is modulated by reward.

I don't think this result can be trusted...

1. The time spent at low-speed at reward sites wasn't controlled. Rats probably left the unrewarded sites much faster, or would move a lot more there than when reward was present.
2. The amount of ripples or replay is analysed in count per second. I believe this means seconds of any kind of behaviour, not just rest or low speed. But #Ripples and the related #Replay only happens when the rat is quasi-immobile. At low speed you might have 1 replay per second, while when moving it will be 0/s.
3. A recent preprint (Mallory et al., 2024) suggests that after a few seconds at a reward location, reverse replay dominates. Ambrose et al. specifically show an increase of reverse replay for those more rewarded locations. If rats indeed spent more time at the most rewarded location, they would be in reverse replay mode, fully explaining the results without the need to invoke any kind of value processing.

I might have missed something (if so let me know) but I am now completely unconvinced by the 2016 paper. What do you all think?

#Hippocampus #ValueCoding